Risk of circulatory diseases associated with proton-pump inhibitors: a retrospective cohort study using electronic medical records in Thailand.

Background: Proton-pump inhibitors (PPIs) are prescribed to treat gastric acid-related diseases, while they may also have potential risks to population health. Recent studies suggested that a potential mechanism explaining the association between PPIs and cardiovascular diseases (CVD) includes the inhibition of the nitrate-nitrite-nitric oxide (NO) pathway. However, previous observational studies showed controversial results of the association. In addition, the inhibition of the NO pathway due to PPIs use may lead to peripheral vascular diseases (PVD); however, none of the studies explore the PPI-PVD association. Therefore, this study aimed to evaluate the association of PPIs with circulatory diseases (CVD, ischemic strokes or IS, and PVD). Methods: We conducted a retrospective hospital-based cohort study from Oct 2010 to Sep 2017 in Songkhla province, Thailand. PPIs and histamine 2-receptor antagonists (H2RAs) prescriptions were collected from electronic pharmacy records, while diagnostic outcomes were retrieved from electronic medical records at Songklanagarind hospital. Patients were followed up with an on-treatment approach. Cox proportional hazard models were applied to measure the association comparing PPIs vs H2RAs after 1:1 propensity-score-matching. Sub-group analysis, multi-bias E-values, and array-based sensitivity analysis for some covariates were used to assess the robustness of associations. Results: A total of 3,928 new PPIs and 3,928 H2RAs users were included in the 1:1 propensity score-matched cohort. As compared with H2RAs, the association of PPIs with CVD, IS, and PVD, the hazard ratios were 1.76 95% CI = [1.40-2.20] for CVD, 3.53 95% CI = [2.21-5.64] for ischemic strokes, and 17.07 95% CI = [13.82-76.25] for PVD. The association between PPIs and each outcome was significant with medication persistent ratio of over 50%. In addition, the association between PPIs and circulatory diseases was robust to unmeasured confounders (i.e., smoking and alcohol). Conclusion: PPIs were associated with circulatory diseases, particularly ischemic strokes in this hospital-based cohort study, whereas, the strength of associations was robust to unmeasured confounders.

Estimates of Chronic Kidney Diseases Associated with Proton-Pump Inhibitors Using a Retrospective Hospital-Based Cohort in Thailand.

Purpose: Potential adverse outcomes of Proton pump inhibitors (PPIs) have increasingly been reported. The potential risks to PPIs include hypomagnesemia and chronic kidney disease (CKD). Unlike a real-world electronic medical record (RW-EMR) with active-comparator design, claim databases and special population cohort with non-user design, using in previous studies, resulted in a wide range of strength of association with indication bias. This study aimed to measure the total effect of association between PPIs use and CKD incidence using Thai RW-EMR. Patients and Methods: A retrospective hospital-based cohort was applied into this study. Electronic medical records and administrative data of out- and inpatient were retrieved from October 1st, 2010 to September 30th, 2017. On-treatment with grace period as well as propensity score matching was used in data analysis. Cox proportional hazard models were applied to evaluate the PPIs-CKD association. Results: Of all 63,595 participants, a total of 59,477 new PPIs and 4118 Histamine 2-receptor antagonist (H2RA) users were eligible for follow-up. As compared with H2RA, the PPI users were non-elderly and more likely being female. The association of PPIs with CKD was statistically significant (adjusted hazard ratio [HR] = 3.753, 95% CI = 2.385-5.905). The HR were not statistically different by concomitant use PPIs with NSAIDs and by medication possession ratio levels. Conclusion: The association between PPIs and CKD incidence was statistically significant in this hospital-based cohort. However, self-treatment with over-the-counter PPIs, as well as, smoking, drinking alcohol and body mass index could not be fully retrieved, affecting the estimation of treatment effect.

Multi-label classification of symptom terms from free-text bilingual adverse drug reaction reports using natural language processing.

Allergic reactions to medication range from mild to severe or even life-threatening. Proper documentation of patient allergy information is critical for safe prescription, avoiding drug interactions, and reducing healthcare costs. Allergy information is regularly obtained during the medical interview, but is often poorly documented in electronic health records (EHRs). While many EHRs allow for structured adverse drug reaction (ADR) reporting, a free-text entry is still common. The resulting information is neither interoperable nor easily reusable for other applications, such as clinical decision support systems and prescription alerts. Current approaches require pharmacists to review and code ADRs documented by healthcare professionals. Recently, the effectiveness of machine algorithms in natural language processing (NLP) has been widely demonstrated. Our study aims to develop and evaluate different NLP algorithms that can encode unstructured ADRs stored in EHRs into institutional symptom terms. Our dataset consists of 79,712 pharmacist-reviewed drug allergy records. We evaluated three NLP techniques: Naive Bayes-Support Vector Machine (NB-SVM), Universal Language Model Fine-tuning (ULMFiT), and Bidirectional Encoder Representations from Transformers (BERT). We tested different general-domain pre-trained BERT models, including mBERT, XLM-RoBERTa, and WanchanBERTa, as well as our domain-specific AllergyRoBERTa, which was pre-trained from scratch on our corpus. Overall, BERT models had the highest performance. NB-SVM outperformed ULMFiT and BERT for several symptom terms that are not frequently coded. The ensemble model achieved an exact match ratio of 95.33%, a F1 score of 98.88%, and a mean average precision of 97.07% for the 36 most frequently coded symptom terms. The model was then further developed into a symptom term suggestion system and achieved a Krippendorff's alpha agreement coefficient of 0.7081 in prospective testing with pharmacists. Some degree of automation could both accelerate the availability of allergy information and reduce the efforts for human coding.