Development and validation of an electronic health records‐based opioid use disorder algorithm by expert clinical adjudication among patients with prescribed opioids

Ranapurwala, Shabbar I.; Alam, Ishrat; Pence, Brian W.; Carey, Timothy S.; Christensen, Sean; Clark, Marshall; Chelminski, Paul R.; Wu, Li-Tzy; Greenblatt, Lawrence H; Korte, Jeffrey E.; Wolfson, Mark; Douglas, Heather; Bowlby, Lynn; Capata, Michael

doi:10.1002/pds.5591

Cited by 6 publications

(6 citation statements)

References 21 publications

(65 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is an attempt to use ML in the broader health system space to advance this field as others have done within smaller systems 42 , 43 and to improve predictive, clinical analytics. 44 , 45 The importance of this work is that it begins to bridge the gap between systems research and clinical research, with a key next step focus on combining ML and cluster analysis at the system level with individual-level claims and/or electronic medical record data—analyses that the authors believe are essential to bridging the gap between clinical outcomes and key SDOH.…”

Section: Discussionmentioning

confidence: 99%

American clusters: using machine learning to understand health and health care disparities in the United States

Bowser,

Maurico,

Ruscitti

et al. 2024

Health Affairs Scholar

View full text Add to dashboard Cite

Health and health care access in the U.S. is plagued by high inequality. While machine learning (ML) is increasingly used in clinical settings to inform health care delivery decisions and predict health care utilization, using ML as a research tool to understand health care disparities in the U.S. and how these are connected to health outcomes, access to health care, and health system organization is less common. We utilized over 650 variables from 24 different databases aggregated by the Agency for Healthcare Research and Quality (AHRQ) in their Social Determinant of Health Database (SDOH). We used k-means–a non-hierarchical ML clustering method–to cluster county level data. Principal factor analysis created county level index values for each SDOH domain and two health care domains–health care infrastructure and health care access. Logistic regression classification was used to identify the primary drivers of cluster classification. The most efficient cluster classification consists of 3 distinct clusters in the U.S.; the cluster having the highest life expectancy comprised only 10% of counties. The most efficient ML clusters do not identify the clusters with the widest health care disparities. ML clustering, using county level data, shows that health care infrastructure and access are the primary drivers of cluster composition.

show abstract

Section: Discussionmentioning

confidence: 99%

American clusters: using machine learning to understand health and health care disparities in the United States

Bowser,

Maurico,

Ruscitti

et al. 2024

Health Affairs Scholar

View full text Add to dashboard Cite

show abstract

“…Fifth, this study relied on ICD codes available in administrative claims data for OUD identification. Because the specificity and sensitivity of OUD ICD codes are unknown and may lead to under-identification [ 47 ], our estimates would be subject to measurement error and could be attenuated. Further, considering this limitation and recognizing the chronic nature of OUD, our findings could be more accurately interpreted as indicating an increased risk of suicide attempts by opioid poisoning, especially among those women diagnosed with OUD both during pregnancy and postpartum.…”

Section: Discussionmentioning

confidence: 99%

Prenatal opioid use as a predictor of postpartum suicide attempts among reproductive-age women enrolled in Oregon Medicaid

Yoon,

Masoumirad,

Bui

et al. 2024

BMC Women's Health

View full text Add to dashboard Cite

Background The rates of suicide and opioid use disorder (OUD) among pregnant and postpartum women continue to increase. This research characterized OUD and suicide attempts among Medicaid-enrolled perinatal women and examined prenatal OUD diagnosis as a marker for postpartum suicide attempts. Methods Data from Oregon birth certificates, Medicaid eligibility and claims files, and hospital discharge records were linked and analyzed. The sample included Oregon Medicaid women aged 15–44 who became pregnant and gave live births between January 2008 and January 2016 (N = 61,481). Key measures included indicators of suicide attempts (separately for any means and opioid poisoning) and OUD diagnosis, separately assessed during pregnancy and the one-year postpartum period. Probit regression was used to examine the overall relationship between prenatal OUD diagnosis and postpartum suicide attempts. A simultaneous equations model was employed to explore the link between prenatal OUD diagnosis and postpartum suicide attempts, mediated by postpartum OUD diagnosis. Results Thirty-three prenatal suicide attempts by any means were identified. Postpartum suicide attempts were more frequent with 58 attempts, corresponding to a rate of 94.3 attempts per 100,000. Of these attempts, 79% (46 attempts) involved opioid poisoning. A total of 1,799 unique women (4.6% of the sample) were diagnosed with OUD either during pregnancy or one-year postpartum with 53% receiving the diagnosis postpartum. Postpartum suicide attempts by opioid poisoning increased from 55.5 per 100,000 in 2009 to 105.1 per 100,000 in 2016. The rate of prenatal OUD also almost doubled over the same period. Prenatal OUD diagnosis was associated with a 0.15%-point increase in the probability of suicide attempts by opioid poisoning within the first year postpartum. This increase reflects a three-fold increase compared to the rate for women without a prenatal OUD diagnosis. A prenatal OUD diagnosis was significantly associated with an elevated risk of postpartum suicide attempts by opioid poisoning via a postpartum OUD diagnosis. Conclusions The risk of suicide attempt by opioid poisoning is elevated for Medicaid-enrolled reproductive-age women during pregnancy and postpartum. Women diagnosed with prenatal OUD may face an increased risk of postpartum suicides attempts involving opioid poisoning.

show abstract

“…Included patients were identified using ICD-10 codes for diagnoses of OUD. Prior studies validating OUD ICD-10 codes have found low (10-40%) sensitivity but high (> 95%) specificity and thus, our study may exclude some patients with OUD (48,49). The PINC AI database provides charge claims only for medications received during inpatient admission.…”

Section: Online Clinical Investigationsmentioning

confidence: 99%

Variation in Use of Medications for Opioid Use Disorder in Critically Ill Patients Across the United States

Feeney,

Law,

Walkey

et al. 2024

Critical Care Medicine

View full text Add to dashboard Cite

Objectives: To describe practice patterns surrounding the use of medications to treat opioid use disorder (MOUD) in critically ill patients. Design: Retrospective, multicenter, observational study using the Premier AI Healthcare Database. Setting: The study was conducted in U.S. ICUs. Patients: Adult (≥ 18 yr old) patients with a history of opioid use disorder (OUD) admitted to an ICU between 2016 and 2020. Interventions: None. Measurements and Main Results: Of 108,189 ICU patients (658 hospitals) with a history of OUD, 20,508 patients (19.0%) received MOUD. Of patients receiving MOUD, 13,745 (67.0%) received methadone, 2,950 (14.4%) received buprenorphine, and 4,227 (20.6%) received buprenorphine/naloxone. MOUD use occurred in 37.9% of patients who received invasive mechanical ventilation. The median day of MOUD initiation was hospital day 2 (interquartile range [IQR] 1–3) and the median duration of MOUD use was 4 days (IQR 2–8). MOUD use per hospital was highly variable (median 16.0%; IQR 10–24; range, 0–70.0%); admitting hospital explained 8.9% of variation in MOUD use. A primary admitting diagnosis of unintentional poisoning (aOR 0.41; 95% CI, 0.38–0.45), presence of an additional substance use disorder (aOR 0.66; 95% CI, 0.64–0.68), and factors indicating greater severity of illness were associated with reduced odds of receiving MOUD in the ICU. Conclusions: In a large multicenter, retrospective study, there was large variation in the use of MOUD among ICU patients with a history of OUD. These results inform future studies seeking to optimize the approach to MOUD use during critical illness.

show abstract

Development and validation of an electronic health records‐based opioid use disorder algorithm by expert clinical adjudication among patients with prescribed opioids

Cited by 6 publications

References 21 publications

American clusters: using machine learning to understand health and health care disparities in the United States

American clusters: using machine learning to understand health and health care disparities in the United States

Prenatal opioid use as a predictor of postpartum suicide attempts among reproductive-age women enrolled in Oregon Medicaid

Variation in Use of Medications for Opioid Use Disorder in Critically Ill Patients Across the United States

Contact Info

Product

Resources

About