Structuring medication signeturs as a language regression task: comparison of zero- and few-shot GPT with fine-tuned models.

Importance: Electronic health record textual sources such as medication signeturs (sigs) contain valuable information that is not always available in structured form. Commonly processed through manual annotation, this repetitive and time-consuming task could be fully automated using large language models (LLMs). While most sigs include simple instructions, some include complex patterns. Objectives: We aimed to compare the performance of GPT-3.5 and GPT-4 with smaller fine-tuned models (ClinicalBERT, BlueBERT) in extracting the average daily dose of 2 immunomodulating medications with frequent complex sigs: hydroxychloroquine, and prednisone. Methods: Using manually annotated sigs as the gold standard, we compared the performance of these models in 702 hydroxychloroquine and 22 104 prednisone prescriptions. Results: GPT-4 vastly outperformed all other models for this task at any level of in-context learning. With 100 in-context examples, the model correctly annotates 94% of hydroxychloroquine and 95% of prednisone sigs to within 1 significant digit. Error analysis conducted by 2 additional manual annotators on annotator-model disagreements suggests that the vast majority of disagreements are model errors. Many model errors relate to ambiguous sigs on which there was also frequent annotator disagreement. Discussion: Paired with minimal manual annotation, GPT-4 achieved excellent performance for language regression of complex medication sigs and vastly outperforms GPT-3.5, ClinicalBERT, and BlueBERT. However, the number of in-context examples needed to reach maximum performance was similar to GPT-3.5. Conclusion: LLMs show great potential to rapidly extract structured data from sigs in no-code fashion for clinical and research applications.

Challenges of caring for homeless patients with inflammatory arthritis: 12-month follow-up observations and identification of certain barriers to care.

Homelessness is a public health crisis and there is a paucity of information about patients with rheumatic disease experiencing homelessness. We sought to develop approaches to improve care for this unique patient population. We previously reported observations on 17 homeless patients with inflammatory arthritis (15 rheumatoid arthritis (RA), 2 psoriatic arthritis (PsA)). We obtained follow-up information from our original 17 patients and compared this to data summarized and published about them from 12 months previously. We also created and administered a 100-question needs assessment survey. Follow-up 12-month clinical information was available from 13/17 homeless and 13/17 non-homeless controls. Homeless patients remained less well with more disease than non-homeless patients-poorer access to clinic appointments (80% vs 91%, p < 0.05), more emergency services use (20 vs 5 ED visits), less DMARDs use (43% vs 100%, p < 0.01), and more steroid use (29% vs 0%, p < 0.01). Homeless patients also had higher inflammatory markers than non-homeless patients (ESR 32 vs 26 mm/h and CRP 17 vs 5 mg/L), although these findings were not statistically significantly different. Seventy-eight percent of homeless patients were stable, 14% improved, and 7% worse; 21% had stable controlled and 57% stable active disease vs 62% and 0% of non-homeless (p < 0.01). Among the homeless, 6 (4 RA, 2 PsA) completed the survey, 2 declined, and 9 could not be reached. All 6 had found housing although all still had housing insecurity; 4 (67%) were homeless in the past. Three out of six (50%) obtained housing from social assistance during hospitalization following disease exacerbation while homeless. The average monthly income was $873. 5/6 (83.3%), were unable to work due to health, and were in considerable pain that adversely impacted their physical and mental health and ability to perform ADLs. Their perceived "greatest need" included dental care, physical therapy, knee surgery, employment, socialization secondary to isolation, and stable housing. Our understanding of the unique challenges of patients with rheumatic disease experiencing homelessness is improved, but not complete. Strengthened collaboration between street medicine providers and rheumatologists is necessary to improve care for homeless patients, especially given poorer outcomes compared with non-homeless counterparts. Key Points • We report 12-month follow-up information from our original 17 homeless patients with inflammatory arthritis (related in this journal in 2021) and their responses to an extensive needs assessment survey designed to identify barriers to care. • Homeless patients with inflammatory arthritis continued to have worse disease outcomes, use more corticosteroids and less DMARDs, and be seen less often in rheumatology clinics and more frequently in emergency departments than their non-homeless counterparts. • Survey data indicated that social assistance during hospitalization was a key area where healthcare providers could intervene to provide housing security for homeless patients and improve outcomes. Patients perceived "greatest needs" went beyond housing and rheumatological care and critically included access to social/specialty services. • Street medicine is the direct delivery of healthcare to people experiencing homelessness wherever they reside. Our observations, obtained in collaboration with street medicine colleagues, suggest important and salutary opportunities for this partnership to improve care for these particular patients.

Coincident Kikuchi-Fujimoto's disease and adult-onset Still's disease: report of a patient from an uncommonly affected population and case-directed systematic review.

Kikuchi-Fujimoto's disease (KFD) and adult-onset Still's disease (AOSD) are rare idiopathic inflammatory conditions of unknown etiology. Ten prior instances of KFD and AOSD occurring together have been reported in the medical literature. These overlaps, together with certain distinguishing clinical and laboratory characteristics in these co-occurrences, offer insight into the pathophysiology of both of these rare disorders. Too, examination of these cases may help improve the diagnostic evaluation and care of patients afflicted with these rare diseases. We therefore report an additional patient with KFD and AOSD occurring in a middle-aged Hispanic female patient and perform a systematic literature review using the PubMed/MEDLINE and Embase databases to further analyze and compare prior identified cases. Our observations in our index case complement and expand previous reports, including new demographic and diagnostic features not seen in prior cases of overlap. Indeed ours is the first in a patient of Hispanic ethnicity, with retroperitoneal lymphadenopathy, as well as with a skin biopsy consistent with AOSD. Each of the reviewed cases of co-occurrence met the diagnostic criteria for both KFD and AOSD. This finding, in the setting of unique clinical and diagnostic manifestations that are not typically seen in either disease entity alone, suggests the presence of an overlap syndrome. Also, many of the shared clinical features and symptomatic responses to targeted therapies implies a similar, yet still poorly understood, pathophysiologic pathway for the two diseases.