Herbal or dietary supplement use is highly prevalent in the U.S. The 2007 National Health Interview Survey (NHIS) reported that 17.9% of adults had used herbal supplements during the previous 12 months. The National Health and Nutrition Examination Survey (NHANES) 2003-2006 reported a much higher estimate of usage: 45% in adults at or below 50 and 67% in adults above 50 years old. The safety of herbal products is a serious concern. Because herbal medicines are not regulated by the U.S. Food and Drug Administration (FDA) before marketing, the adverse effects and drug interactions associated with many products are not well known. In addition, adverse reactions to the products are often not reported. Meanwhile, many herb-drug-disease interactions have been reported in the literature. Partially because of the effective surveillance of prescription drug-drug interaction, a large survey study found that the majority of potential interactions involved herbal supplements. Studies point to the need to improve the surveillance and prevention of herb-drug interaction.
Our long-term goal is to improve the surveillance and prevention of herb-drug interactions in the VHA population. The goal of this pilot study is to assess the feasibility of using EHR data to estimate exposure to dietary supplements.
We will first survey veterans' current and past herbal supplement usage. Questionnaire results will be observed as the gold standard, and we will identify documentation of herbal supplement use in the clinical notes and structured fields of the VA EHR. The questionnaire will focus on current and past use of herbal supplements; specific products taken, frequency/duration, and whether or not physicians were informed about supplement use. We will concentrate on two commonly used herbal supplements: Gingko and St John's Wort. Clinical notes will be retrieved from the VA EMR, in order to establish the degree to which data in the chart accurately reflects patient usage. Finally, we will use NLP tools to develop and validate an automated method to identify herbal supplement(s) exposure based on EHR data.
We surveyed 377 patients from 3 VAMCs (130 from DCVA, 132 from SLC VA and 115 from West Haven). The demographics for: (1) DC participants are- Age: Min 32, Mean 61.5, Max 88; Gender: 20(20%) female, 110 (85%) male, and 0 unknown; Race: 105 (81%) black, 16 (12%) white, 1 (1%) other and 8 (6%) unknown; Ethnicity: 1(1%) Hispanic, 123 (94%) Non-Hispanic, and 6 (5%) unknown. (2) SLC participants are: Age - Min 24, Mean 64.7, Max 93; Gender: 7 (5%) female, 125 (95%) male, and 0 unknown; Race: 9 (7%) black, 115 (87%) white, 8 (6%) other and 0 unknown; Ethnicity: 9 (7%) Hispanic, 123 (93%) Non-Hispanic, and 0 unknown. (3) West Heaven participants are: Age - Min 32, Mean 59.1, Max 88; Gender: 13 (11%) female, 101 (88%) male, and 1 (1%) unknown; Race: 41 (36%) black, 67 (58%) white, 2 (2%) other and 5 (4%) unknown; Ethnicity: 3 (3%) Hispanic, 107 (93%) Non-Hispanic, and 5 (4%) unknown.
Among them, 288 (76%) took at least 1 dietary supplement in the past 12 months. Among those who took supplement, 77% told their providers. Most of the disclosure was made to physicians (91%), followed by nurse (24%) and pharmacist (12%). Common supplements used were Multi-Vitamin, Vitamin B, C, and D, fish oil, garlic, green tea and melatonin. We started out focusing on just herbal products but decided to broaden the scope (mainly, including minerals and vitamins along with herbs) because many patients do not distinguish the types of supplements. A detailed breakdown of supplements used by participants (from DC, SLC and West Haven, respectively) is as follows: .Aloe Vera (13, 0, 0); Black Cohosh (2,0,0 ); Calcium (0,7,7); Chondroitin (2,8,7); Cinnamon (11,0, 2); Coenzyme Q (2,6,7); Cranberry (18,0,0); DHEA (1,1,5); Echinacea (3,2,2); Evening Primrose (1,0,0); Fish Oil (31,13,9); Flaxseed (1,1,1); Folic Acid (10,1,6); Garlic (23,15,7); Ginger(13,1,1); Gingko Biloba (7,6,10); Ginseng (10,6,11); Glucosamine (3,17,10); Goldenseal (3,0,0); Green Tea (41,1,2), Iron (0,3,5); Magnesium (0,8,4); Melatonin (3,15,22); Multivitamins/Vitamin (45,49,49); Potassium (0,2,1); Protein (0,1,5); Saw Palmetto (2,2,2); St John's wort (1,1,1);Turmeric (6,1,2); Vitamin A (0,1,3); Vitamin B (27,22,32); Vitamin C (30,9,10); Vitamin D (40,27,39); Vitamin E (21,7,2); Yohimbe (2,0,0) and Zinc (0,1,2).
We first developed a misspelling detection algorithm and applied to the discovery of misspelled supplement terms. The algorithm relies on a word2vect model trained using a large corpus (approximately 1.6 million records), achieving positive predictive value of 93%. Using the patient reported keywords and potential misspells, we extracted a large number of snippets. We randomly sampled 1000 snippets to annotate. The NLP module trained on the annotated snippets achieved 90% F measure.
Ginkgo biloba is a widely used herbal product that could potentially have a severe interaction with warfarin, which is the most frequently prescribed anticoagulant agent in North America. Literature, however, provides conflicting evidence on the presence and severity of the interaction. We developed NLP methods to extract the ginkgo usage and combined it with prescription data on warfarin from a very large clinical data repository. Our statistical analysis suggests that taking concurrently with warfarin, gingko does significantly increase patients' risk of a bleeding adverse event (hazard ratio = 1.38, 95%CI: 1.20 to 1.58, p<.001). This study also is the first attempt of using a large medical record database to confirm a suspected herb-drug interaction.
We identified patients who had serum magnesium test were free of DM at time of their first magnesium test (n=2,973,000). Of these, 565,000 developed new-onset DM during an average follow-up of ~6 years. In preliminary analysis, patients with low magnesium had a higher, raw incident rate of DM (21% vs 18%). Among those with low magnesium, patients who received single compound magnesium supplement had lower incident rate than those with no evidence of magnesium exposure (19% vs. 27%). In addition, we performed a simple logistic regression analysis and included only age, gender, ethnicity, and Charlson Comorbidity Index as co-variates, we found that magnesium supplement use was associated with a lower risk (Odds ratio = 0.70).
The co-administration of herbal supplements and prescription drugs has become a serious concern. In the Veterans Health Administration (VHA), a number of studies have found a high prevalence of herbal supplement usage and frequent co-administration of herb and prescription drugs. Most of these studies are local surveys that have identified potential herb-drug interactions rather than actual adverse events. Since the VHA is a national leader in the use of electronic health records, we have a potential opportunity to study the prevalence and severity of herb-drug interactions by leveraging currently available VHA EHR data, which is a large national dataset. A first step in the use of EHR data to study herb-drug interactions is an analysis of the completeness and accuracy of documented herbal usage in VHA EHR. Our long-term goal is to improve the surveillance and prevention of herb-drug interactions in the VHA population. The goal of this pilot study was to assess the feasibility of using EHR data to estimate exposure to herbal supplements. Our results showed that a majority of Veterans have taken supplements and inform the providers of the use. In addition, we demonstrated that we are able to extract the supplement use from clinical notes with good performance. The study paved the way to our next step which is to analyze supplement-drug interaction and assess potential harms and benefits of supplement use.
None at this time.
Research Infrastructure, Epidemiology, Treatment - Observational
Complementary and Alternative Practices, Natural Language Processing