Classifying Pseudogout Using Machine Learning Approaches With Electronic Health Record Data

Abstract

Objective Identifying pseudogout in large data sets is difficult due to its episodic nature and a lack of billing codes specific to this acute subtype of calcium pyrophosphate (CPP) deposition disease. The objective of this study was to evaluate a novel machine learning approach for classifying pseudogout using electronic health record (EHR) data. Methods We created an EHR data mart of patients with ≥1 relevant billing code or ≥2 natural language processing (NLP) mentions of pseudogout or chondrocalcinosis, 1991–2017. We selected 900 subjects for gold standard chart review for definite pseudogout (synovitis + synovial fluid CPP crystals), probable pseudogout (synovitis + chondrocalcinosis), or not pseudogout. We applied a topic modeling approach to identify definite/probable pseudogout. A combined algorithm included topic modeling plus manually reviewed CPP crystal results. We compared algorithm performance and cohorts identified by billing codes, the presence of CPP crystals, topic modeling, and a combined algorithm. Results Among 900 subjects, 123 (13.7%) had pseudogout by chart review (68 definite, 55 probable). Billing codes had a sensitivity of 65% and a positive predictive value (PPV) of 22% for pseudogout. The presence of CPP crystals had a sensitivity of 29% and a PPV of 92%. Without using CPP crystal results, topic modeling had a sensitivity of 29% and a PPV of 79%. The combined algorithm yielded a sensitivity of 42% and a PPV of 81%. The combined algorithm identified 50% more patients than the presence of CPP crystals; the latter captured a portion of definite pseudogout and missed probable pseudogout. Conclusion For pseudogout, an episodic disease with no specific billing code, combining NLP, machine learning methods, and synovial fluid laboratory results yielded an algorithm that significantly boosted the PPV compared to billing codes.

Proton‐Pump Inhibitors and Risk of Calcium Pyrophosphate Deposition in a Population‐Based Study

“…While use of serial radiographs to capture truly incident chondrocalcinosis is possible in a large observational cohort followed longitudinally, this remains a challenge in large databases. To overcome some of these limitations for case ascertainment, Tedeschi et al developed a machine learning–based algorithm to identify CPPD cases in the EMR (33). Using this algorithm, PPI use was associated with a 2‐fold higher odds of the diagnosis of pseudogout, supporting the findings of prior studies.…”

Confirming Prior and Identifying Novel Correlates of Acute Calcium Pyrophosphate Crystal Arthritis

“…Misclassification of patients with other types of arthritis or other CPPD manifestations such as chronic CPP crystal arthritis was expected to occur in approximately 19% of patients per our algorithm (PPV of 81%). As previously reported, we manually reviewed 100 randomly selected cases; 85 had acute CPP crystal arthritis, 9 had osteoarthritis with CPPD, 4 had possible acute CPP crystal arthritis, and 2 had crystal-proven gout (7). We thus expect that only a very small percentage of gout patients were misclassified as having acute CPP crystal arthritis.…”

“…Well‐characterized, large cohorts of acute CPP crystal arthritis patients are necessary for confirming associated conditions and investigating long‐term outcomes of this inflammatory crystalline arthritis. We previously developed an algorithm to accurately identify acute CPP crystal arthritis using electronic health record (EHR) data (7). The algorithm leverages information captured in narrative notes, radiology reports, and laboratory data, including synovial fluid crystal analysis, and other structured and unstructured EHR data to achieve a positive predictive value (PPV) of 81%, notably higher than the PPV of 22% for acute CPP crystal arthritis using billing codes alone (7).…”

Risk of cardiovascular events in patients having had acute calcium pyrophosphate crystal arthritis

“…The acute CPP crystal arthritis cohort was identified by applying a previously published EHR-based algorithm with positive predictive value (PPV) 81% for this acute inflammatory manifestation of CPPD 16. The algorithm uses machine learning techniques to classify patients as acute CPP crystal arthritis or not, incorporating a range of EHR information including natural language processing of narrative notes and radiology reports, laboratory data including synovial fluid crystal analysis, and structured EHR data such as billing codes.…”