Using natural language processing to identify problem usage of prescription opioids

The prevalence of problem opioid use in patients receiving chronic opioid therapy

“…Second, there was a population who had a diagnosis code which had been carried forward through time but who did not have documentation of current problem opioid use. 7 Consistent with previous research, 6,12,20 we found higher rates of problem opioid use among younger patients. Although males had slightly higher prevalence rates of problem opioid use than females, our results were consistent with previous research suggesting that gender differences in rates of problem opioid use are modest.…”

“…The reasons for discrepancies between ICD-9 codes and NLPidentified problems with opioid use are considered in depth in a separate article reporting the development of the NLP algorithms. 7 Common reasons for divergence included physiological dependence in and of itself was denoted with an ICD-9 code but there was no clinical documentation of problem opioid use, previous diagnoses of opioid abuse or dependence were carried forward to current clinical notes but without documentation of current problem opioid use, and circumstances that suggested potential problem opioid use accompanied by the diagnostic code, but without sufficient documentation to identify opioid problems from the clinical notes based on the criteria summarized in Table 1. The prevalence rates for specific ICD-9 codes are presented in Table 2.…”

“…7 A custom dictionary of terms was developed that described the common manifestations of problem opioid use such as opioid abuse, misuse, overuse, addiction, and diversion. This dictionary was developed by the authors in consultation with the practicing physicians within Group Health.…”

Use of Opioid Agonist Therapy for Medicare Patients in 2013

“…3 We believe this reflects a significant treatment gap, although we are limited in providing precise estimates; not all patients with an opioid use disorder warrant OAT, but on the other hand, opioid disorders are systematically under diagnosed and increasing in prevalence. 5 Furthermore, more than one-third of Part D enrollees fill at least 1 prescription for an opioid in any given year, 3 putting many more patients at risk for iatrogenic addiction. 6 …”

Using prescription claims to detect aberrant behaviors with opioids: comparison and validation of 5 algorithms

“…These conditions were identified using the inpatient and outpatient International Classification of Diseases, 9th Revision, Clinical Modification codes (detailed definitions in Appendix Table A2). Previous validation studies suggest a high concordance between International Classification of Diseases, 9th Revision, Clinical Modification codes for opioid abuse and dependence and overdose and diagnoses for adverse opioid events recorded in medical records . In sensitivity analyses to ensure high specificity of the outcome measure, we further restricted the definition to opioid poisoning alone, which has a positive predictive value of 96% for opioid‐related adverse events …”