County-Level Atrazine Use and Gastroschisis.

Importance: The incidence of gastroschisis, a birth defect involving the herniation of the small bowel through the abdominal wall, has increased in the US since the 1960s. The pesticide atrazine is a hypothesized cause of gastroschisis; however, examination of the association between atrazine and gastroschisis has been limited. Objective: To evaluate national trends in gastroschisis incidence, maternal and infant characteristics associated with gastroschisis, and whether county-level atrazine use is associated with gastroschisis. Design, Setting, and Participants: This retrospective, repeated cross-sectional study examined birth certificate data of all live births in the US and data on atrazine use from the US Geological Survey from January 1, 2009, through December 31, 2019. The data analysis was performed between August 5, 2021, and May 26, 2023. Exposures: County-level atrazine use. Main Outcomes and Measures: The primary outcome was gastroschisis incidence. Covariates included maternal age, race and ethnicity, body mass index (measured by weight in kilograms divided by height in meters squared), parity, insurance type, Chlamydia infection during pregnancy, smoking, and rurality. Mixed-effects logistic regression models (year fixed effects and county random effects) were constructed using different county-level atrazine exposure variables (1-, 5-, and 10-year means). Results: Between 2009 and 2019, 39 282 566 live births were identified, with 10 527 infant diagnoses of gastroschisis. Infants with gastroschisis were more likely to have mothers who identified as non-Hispanic White (61% vs 54%; P < .001), had a lower body mass index (median [IQR], 23.4 [20.8-27.2] vs 25.4 [22.0-30.8]; P < .001), were more likely to be nulliparous (median [IQR], 0 [0-1] vs 1 [0-2]; P < .001), and were more commonly covered by Medicaid (63% vs 43%; P < .001). During the study period, the rate (per 1000 live births) of gastroschisis decreased from 0.31 (95% CI, 0.29-0.33) to 0.22 (95% CI, 0.21-0.24). The median (IQR) county-level atrazine use estimates were higher among infants with gastroschisis (1 year, 1389 [IQR, 198-10 162] vs 1023 [IQR, 167-6960] kg; 5 years, 1425 [IQR, 273-9895] vs 1057 [IQR, 199-6926] kg; 10 years, 1508 [IQR, 286-10 271] vs 1113 [IQR, 200-6650] kg; P < .001). In adjusted models, higher county levels of atrazine (each 100 000-kg increase) were associated with a higher incidence of gastroschisis (1 year: adjusted odds ratio [AOR], 1.12 [95% CI, 1.01-1.24]; 5 years: AOR, 1.15 [95% CI, 1.02-1.30]; 10 years: AOR, 1.21 [95% CI, 1.07-1.38]). Conclusions and Relevance: In this cross-sectional study, higher county levels of atrazine were associated with infant diagnoses of gastroschisis. While atrazine is the second-most used herbicide in the US, numerous countries around the world have banned it out of concern for adverse effects on human health. These findings suggest that exploring alternatives to atrazine in the US may be warranted.

Buprenorphine Out-of-Pocket Costs and Discontinuation in Privately Insured Adults With Opioid Use Disorder.

This cohort study examined the association between out-of-pocket costs for an initial buprenorphine prescription and its discontinuation among commercially insured US adults with opioid use disorder.

Patient and Community Factors Affecting Treatment Access for Opioid Use Disorder.

OBJECTIVE: To examine whether access to treatment for women with opioid use disorder (OUD) varied by race and ethnicity, community characteristics, and pregnancy status. METHODS: We conducted a secondary data analysis of a simulated patient caller study of buprenorphine-waivered prescribers and opioid-treatment programs in 10 U.S. states. We conducted multivariable analyses, accounting for potential confounders, to evaluate factors associated with likelihood of successfully securing an appointment. Descriptive statistics and significance testing examined 1) caller characteristics and call outcome by assigned race and ethnicity and clinic type (combined, opioid-treatment programs, and buprenorphine-waivered prescribers) and 2) clinic and community characteristics and call outcome by community race and ethnicity distribution (majority White vs majority Black, Hispanic, Asian, American Indian, Alaska Native, Native Hawaiian, or Pacific Islander) and clinic type. A multiple logistic regression model was fitted to assess the likelihood of obtaining an appointment by callers' race and ethnicity and pregnancy status with the exposure of interest being majority Black, Hispanic, Asian, American Indian, Alaska Native, Native Hawaiian, or Pacific Islander community distribution. RESULTS: In total, 3,547 calls reached clinics to schedule appointments. Buprenorphine-waivered prescribers were more likely to be in communities that were more than 50% White (88.9% vs 77.3%, P<.001), and opioid-treatment programs were more likely to be in communities that were less than 50% White (11.1% vs 22.7%, P<.001). Callers were more likely to be granted appointments in majority Black, Hispanic, Asian, American Indian, Alaska Native, Native Hawaiian, or Pacific Islander communities (adjusted odds ratio [aOR] 1.06, 95% CI 1.02-1.10 per 10% Black, Hispanic, Asian, American Indian, Alaska Native, Native Hawaiian, or Pacific Islander community population) and at opioid-treatment programs (aOR 4.94, 95% CI 3.52-6.92) and if they were not pregnant (aOR 1.79, 95% CI 1.53-2.09). CONCLUSION: Clinic distribution and likelihood of acceptance for treatment varied by community race and ethnicity distribution. Access to treatment for OUD remains challenging for pregnant people and in many historically marginalized U.S. communities.

County-level Factors and Treatment Access Among Insured Women With Opioid Use Disorder.

BACKGROUND: An over 40% increase in overdose deaths within the past 2 years and low levels of engagement in treatment call for a better understanding of factors that influence access to medication for opioid use disorder (OUD). OBJECTIVE: To examine whether county-level characteristics influence a caller's ability to secure an appointment with an OUD treatment practitioner, either a buprenorphine-waivered prescriber or an opioid treatment program (OTP). RESEARCH DESIGN AND SUBJECTS: We leveraged data from a randomized field experiment comprised of simulated pregnant and nonpregnant women of reproductive age seeking treatment for OUD among 10 states in the US. We employed a mixed-effects logistic regression model with random intercepts for counties to examine the relationship between appointments received and salient county-level factors related to OUD. MEASURES: Our primary outcome was the caller's ability to secure an appointment with an OUD treatment practitioner. County-level predictor variables included socioeconomic disadvantage rankings, rurality, and OUD treatment/practitioner density. RESULTS: Our sample comprised 3956 reproductive-aged callers; 86% reached a buprenorphine-waivered prescriber and 14% an OTP. We found that 1 additional OTP per 100,000 population was associated with an increase (OR=1.36, 95% CI: 1.08 to 1.71) in the likelihood that a nonpregnant caller receives an OUD treatment appointment from any practitioner. CONCLUSIONS: When OTPs are highly concentrated within a county, women of reproductive age with OUD have an easier time securing an appointment with any practitioner. This finding may suggest greater practitioners' comfort in prescribing when there are robust OUD specialty safety nets in the county.

Treatment access for opioid use disorder among women with medicaid in Florida.

INTRODUCTION: Opioid use disorder (OUD) remains prevalent. Medications for OUD (MOUD) are standard care for pregnant and non-pregnant women. Previous research has identified barriers to MOUD for women with Medicaid but did not account for the type of MOUD (methadone vs. buprenorphine) or pregnancy status. We examined access to MOUD by treatment type for pregnant and non-pregnant women with Medicaid in Florida. METHODS: A secondary analysis of Florida "secret-shopper" data was conducted. Calls were made to clinicians from the 2018 Substance Abuse and Mental Health Services Administration provider list by women posing as either a pregnant or non-pregnant woman with OUD and Medicaid. We examined 546 calls to buprenorphine-waivered providers (BWP) and 139 to opioid treatment programs (OTP). Counts and percentages were used to describe caller success by type of treatment and pregnancy status. Chi-square tests were used to identify statistical differences. RESULTS: Only 42 % of calls reached a treatment provider in Florida. Pregnant and non-pregnant women were less likely to obtain an appointment with Medicaid coverage by a BWP than an OTP (p < 0.01). Nearly 40 % of OTPs offered appointments to callers with Medicaid compared to only 17 % of BWPs. Both types of providers denied appointments more often for pregnant women. Thirty-eight percent of BWP's and 12 % of OTP's denied appointments to pregnant women using cash or Medicaid payment. CONCLUSIONS: Our study demonstrates logistical and financial barriers to treatment for OUD among pregnant and non-pregnant women with Medicaid in Florida and highlights the need for improved systems of care.

Laws Forbidding Pregnancy Discrimination in Substance Use Disorder Treatment Are Not Associated With Treatment Access.

OBJECTIVES: To determine if state laws that forbid substance use disorder treatment providers from refusing services based on pregnancy are associated with improved ability to secure an initial appointment for opioid use disorder treatment with methadone and buprenorphine. METHODS: Ten states in which a "secret shopper" study was conducted were stratified based on whether they have a law that forbids substance use disorder treatment providers from refusing to treat individuals because of pregnancy. We tested for differences in percent of pregnant versus nonpregnant callers offered an initial appointment for medication for opioid use disorder treatment between states with a law and those without. We also tested for differences in acceptance based on caller insurance type. RESULTS: Four states with 1461 callers had relevant laws; 6 states with 2495 callers did not. None of the laws apply to all providers. In both groups of states, nonpregnant callers were significantly more likely to obtain a buprenorphine appointment (75.1% vs 60.1% in states with a law, P < 0.01; 73.1% vs 62.3% in states without, P < 0.01). No significant differences were found in methadone appointment access either between pregnant and nonpregnant callers or between pregnant callers in states with a law compared to those without. CONCLUSIONS: Laws that forbid providers from refusing to treat potential patients because they are pregnant were not associated with success in obtaining an initial appointment for medication for opioid use disorder treatment. States should consider expanding those laws to apply to all treatment providers and ensure that currently covered providers are in compliance.

Moving beyond the classic difference-in-differences model: a simulation study comparing statistical methods for estimating effectiveness of state-level policies.

BACKGROUND: Reliable evaluations of state-level policies are essential for identifying effective policies and informing policymakers' decisions. State-level policy evaluations commonly use a difference-in-differences (DID) study design; yet within this framework, statistical model specification varies notably across studies. More guidance is needed about which set of statistical models perform best when estimating how state-level policies affect outcomes. METHODS: Motivated by applied state-level opioid policy evaluations, we implemented an extensive simulation study to compare the statistical performance of multiple variations of the two-way fixed effect models traditionally used for DID under a range of simulation conditions. We also explored the performance of autoregressive (AR) and GEE models. We simulated policy effects on annual state-level opioid mortality rates and assessed statistical performance using various metrics, including directional bias, magnitude bias, and root mean squared error. We also reported Type I error rates and the rate of correctly rejecting the null hypothesis (e.g., power), given the prevalence of frequentist null hypothesis significance testing in the applied literature. RESULTS: Most linear models resulted in minimal bias. However, non-linear models and population-weighted versions of classic linear two-way fixed effect and linear GEE models yielded considerable bias (60 to 160%). Further, root mean square error was minimized by linear AR models when we examined crude mortality rates and by negative binomial models when we examined raw death counts. In the context of frequentist hypothesis testing, many models yielded high Type I error rates and very low rates of correctly rejecting the null hypothesis (< 10%), raising concerns of spurious conclusions about policy effectiveness in the opioid literature. When considering performance across models, the linear AR models were optimal in terms of directional bias, root mean squared error, Type I error, and correct rejection rates. CONCLUSIONS: The findings highlight notable limitations of commonly used statistical models for DID designs, which are widely used in opioid policy studies and in state policy evaluations more broadly. In contrast, the optimal model we identified--the AR model--is rarely used in state policy evaluation. We urge applied researchers to move beyond the classic DID paradigm and adopt use of AR models.

Neonatal Opioid Withdrawal Syndrome Is Associated With Infant Foster Care Entry At The County Level.

Although the number of infants diagnosed with neonatal opioid withdrawal syndrome (NOWS) and the number of infants entering foster care have increased substantially in the US since 2009, analyses exploring their relationship are lacking. Using data from 580 US counties in eight US states from the period 2009-17, we examined the association of county rates of NOWS and county-level characteristics with infant foster care entries. In adjusted analyses, every one diagnosis of NOWS per ten births was associated with a 41 percent higher rate of infant foster care entry, and rural county residence was associated with a 19 percent higher rate of infant foster entry. A higher employment rate was associated with lower rates of infant foster care entry both overall and in urban counties when we stratified by rurality. These findings suggest that policy makers could use information about county characteristics to better target funding to support opioid-affected families at risk for foster care involvement.

Increasing Volume-Targeted Ventilation Use in the NICU.

BACKGROUND: In preterm infants who require mechanical ventilation (MV), volume-targeted ventilation (VTV) modes are associated with lower rates of bronchopulmonary dysplasia compared with pressure-limited ventilation. Bronchopulmonary dysplasia rates in our NICU were higher than desired, prompting quality improvement initiatives to improve MV by increasing the use of VTV. METHODS: We implemented and tested interventions over a 3-year period. Primary outcomes were the percentage of conventional MV hours when any-VTV mode was used and the percentage of conventional MV hours when an exclusively VTV mode was used. Exclusively VTV modes were modes in which all breaths were volume targeted. We evaluated outcomes during 3 project periods: baseline (May 2016-December 2016); epoch 1 (December 2016-October 2018), increasing the use of any-VTV mode; and epoch 2 (October 2018-November 2019), increasing the use of exclusively VTV modes. RESULTS: Use of any-VTV mode increased from 18 694 of 22 387 (83%) MV hours during baseline to 72 846 of 77 264 (94%) and 58 174 of 60 605 (96%) MV hours during epochs 1 and 2, respectively (P < .001). Use of exclusively VTV increased from 5967 of 22 387 (27%) during baseline to 47 364 of 77 264 (61%) and 46 091 of 60 605 (76%) of all conventional MV hours during epochs 1 and 2, respectively (P < .001). In statistical process control analyses, multiple interventions were associated with improvements in primary outcomes. Measured clinical outcomes were unchanged. CONCLUSIONS: Quality improvement interventions were associated with improved use of VTV but no change in measured clinical outcomes.

Association of Individual and Community Factors With Hepatitis C Infections Among Pregnant People and Newborns.

Importance: The opioid crisis has increasingly affected pregnant people and infants. Hepatitis C virus (HCV) infections, a known complication of opioid use, grew in parallel with opioid-related complications; however, the literature informing individual and community risks associated with maternal HCV infection is sparse. Objectives: To determine (1) individual and county-level factors associated with HCV among pregnant people and their newborn infants, and (2) how county-level factors influence individual risk among the highest risk individuals. Design Setting and Participants: This time-series analysis of retrospective, repeated cross-sectional data included pregnant people in all US counties from 2009 to 2019. We constructed mixed-effects logistic regression models to explore the association between HCV infection and individual and county-level covariates. Analyses were conducted between June 2019 and September 2021. Exposures: Individual-level: race and ethnicity, education, marital status, insurance type; county-level: rurality, employment, density of obstetricians. Main Outcomes and Measures: Hepatitis C virus as indicated on the newborn's birth certificate. Results: Between 2009 and 2019, there were 39 380 122 pregnant people who met inclusion criteria, among whom 138 343 (0.4%) were diagnosed with HCV. People with HCV were more likely to be White (79.9% vs 53.5%), American Indian or Alaska Native (AI/AN) (2.9% vs 0.9%), be without a 4-year degree (93.2% vs 68.6%), and be unmarried (73.7% vs 38.8%). The rate (per 1000 live births) of HCV among pregnant people increased from 1.8 to 5.1. In adjusted analyses, the following factors were associated with higher rates of HCV: individuals identified as White (adjusted odds ratio [aOR], 7.37; 95% CI, 7.20-7.55) and AI/AN (aOR, 7.94; 95% CI, 7.58-8.31) compared with Black individuals, those without a 4-year degree (aOR, 3.19; 95% CI, 3.11-3.28), those with Medicaid vs private insurance (aOR, 3.27; 95% CI, 3.21-3.33), and those who were unmarried (aOR, 2.80; 95% CI, 2.76-2.84); whereas, rural residence, higher rates of employment, and greater density of obstetricians was associated with lower risk of HCV. Among individuals at the highest risk of HCV, higher levels of county employment, accounting for other factors, were associated with less of a rise in HCV infections over time. Conclusions and Relevance: In this cross-sectional study, maternal and newborn HCV infections increased substantially between 2009 and 2019, disproportionately among White and AI/AN people without a 4-year degree. County-level factors, including higher levels of employment, were associated with lower individual risks of acquiring the virus.

Evaluating Definitions for Neonatal Abstinence Syndrome.

BACKGROUND AND OBJECTIVES: National estimates indicate that the incidence of neonatal abstinence syndrome (NAS), a postnatal opioid withdrawal syndrome, increased more than fivefold between 2004 and 2016. There is no gold standard definition for capturing NAS across clinical, research, and public health settings. Our objective was to evaluate how different definitions of NAS modify the calculated incidence when applied to a known population of opioid-exposed infants. METHODS: Data for this retrospective cohort study were obtained from opioid-exposed infants born at Vanderbilt University Medical Center in 2018. Six commonly used clinical and surveillance definitions of opioid exposure and NAS were applied to the study population and evaluated for accuracy in assessing clinical withdrawal. RESULTS: A total of 121 opioid-exposed infants met the criteria for inclusion in our study. The proportion of infants who met criteria for NAS varied by predefined definition, ranging from 17.4% for infants who received morphine to 52.8% for infants with the diagnostic code for opioid exposure. Twenty-eight infants (23.1%) received a clinical diagnosis of NAS by a medical provider, and 38 (34.1%) received the diagnostic code for NAS at discharge. CONCLUSIONS: We found significant variability in the incidence of opioid exposure and NAS among a single-center population using 6 common definitions. Our findings suggest a need to develop a gold standard definition to be used across clinical, research, and public health surveillance settings.

Building longitudinal medication dose data using medication information extracted from clinical notes in electronic health records.

OBJECTIVE: To develop an algorithm for building longitudinal medication dose datasets using information extracted from clinical notes in electronic health records (EHRs). MATERIALS AND METHODS: We developed an algorithm that converts medication information extracted using natural language processing (NLP) into a usable format and builds longitudinal medication dose datasets. We evaluated the algorithm on 2 medications extracted from clinical notes of Vanderbilt's EHR and externally validated the algorithm using clinical notes from the MIMIC-III clinical care database. RESULTS: For the evaluation using Vanderbilt's EHR data, the performance of our algorithm was excellent; F1-measures were ≥0.98 for both dose intake and daily dose. For the external validation using MIMIC-III, the algorithm achieved F1-measures ≥0.85 for dose intake and ≥0.82 for daily dose. DISCUSSION: Our algorithm addresses the challenge of building longitudinal medication dose data using information extracted from clinical notes. Overall performance was excellent, but the algorithm can perform poorly when incorrect information is extracted by NLP systems. Although it performed reasonably well when applied to the external data source, its performance was worse due to differences in the way the drug information was written. The algorithm is implemented in the R package, "EHR," and the extracted data from Vanderbilt's EHRs along with the gold standards are provided so that users can reproduce the results and help improve the algorithm. CONCLUSION: Our algorithm for building longitudinal dose data provides a straightforward way to use EHR data for medication-based studies. The external validation results suggest its potential for applicability to other systems.

Association of Pregnancy and Insurance Status With Treatment Access for Opioid Use Disorder.

Importance: Medications for opioid use disorder, including buprenorphine hydrochloride and methadone hydrochloride, are highly effective at improving outcomes for individuals with the disorder. For pregnant women, use of these medications also improves pregnancy outcomes, including the risk of preterm birth. Despite the known benefits of medications for opioid use disorder, many pregnant and nonpregnant women with the disorder are not receiving them. Objective: To determine whether pregnancy and insurance status are associated with a woman's ability to obtain an appointment with an opioid use disorder treatment clinician. Design, Setting, and Participants: In this cross-sectional study with random assignment of clinicians and simulated-patient callers (performed in "secret shopper" format), outpatient clinics that provide buprenorphine and methadone were randomly selected from publicly available treatment lists in 10 US states (selected for variability in opioid-related outcomes and policies) from March 7 to September 5, 2019. Pregnant vs nonpregnant woman and private vs public insurance assigned randomly to callers to create unique patient profiles. Simulated patients called the clinics posing as pregnant or nonpregnant women to obtain an initial appointment with a clinician. Main Outcomes and Measures: Appointment scheduling, wait time, and out-of-pocket costs. Results: A total of 10 871 unique patient profiles were assigned to 6324 clinicians. Among all women, 2312 of 3420 (67.6%) received an appointment with a clinician who prescribed buprenorphine, with lower rates among pregnant vs nonpregnant callers (1055 of 1718 [61.4%] vs 1257 of 1702 [73.9%]; relative risk, 0.83; 95% CI, 0.79-0.87). For clinicians who prescribed methadone, there was no difference in appointment access for pregnant vs nonpregnant callers (240 of 271 [88.6%] vs 237 of 265 [89.4%]; relative risk, 0.99; 95% CI, 0.93-1.05). Insurance was frequently not accepted, with 894 of 3420 buprenorphine-waivered prescribers (26.1%) and 174 of 536 opioid treatment programs (32.5%) granting appointments only when patients agreed to pay cash. Median wait times did not differ between pregnant and nonpregnant callers among buprenorphine prescribers (3 days [interquartile range, 1-7 days] vs 3 days [interquartile range, 1-7 days]; P = .43) but did differ among methadone prescribers (1 day [interquartile range, 1-4 days] vs 2 days [interquartile range, 1-6 days]; P = .049). For patients agreeing to pay cash, the median out-of-pocket costs for initial appointments were $250 (interquartile range, $155-$300) at buprenorphine prescribers and $34 (interquartile range, $15-$120) at methadone prescribers. Conclusions and Relevance: In this cross-sectional study with random assignment of clinicians and simulated-patient callers, many women, especially pregnant women, faced barriers to accessing treatment. Given the high out-of-pocket costs and lack of acceptance of insurance among many clinicians, access to affordable opioid use disorder treatment is a significant concern.

Neonatal Abstinence Syndrome In The United States, 2004-16.

After increasing for nearly two decades, rates of neonatal abstinence syndrome have recently leveled off, reaching a plateau as early as 2014. These findings may represent successful efforts to prevent and treat opioid use before and during pregnancy.

Hepatitis C Testing Among Perinatally Exposed Infants.

BACKGROUND: Hepatitis C virus (HCV) prevalence doubled among pregnant women from 2009 to 2014, reaching 3.4 per 1000 births nationwide. Infants exposed to HCV may acquire HCV by vertical transmission. National guidelines recommend that infants exposed to HCV be tested; however, it is unclear if these recommendations are being followed. Our objectives were to determine if infants exposed to HCV were tested and to determine hospital- and patient-level factors associated with differences in testing. METHODS: In this retrospective cohort study of infants exposed to HCV who were enrolled in the Tennessee Medicaid program, we used vital statistics-linked administrative data for infants born between January 1, 2005, and December 31, 2014. Infants were followed until 2 years old. Multilevel logistic regression was used to assess the association of HCV testing and hospital- and patient-level characteristics. RESULTS: Only 23% of 4072 infants exposed to HCV were tested. Infants whose mothers were white versus African American (96.6% vs 3.1%; P <.001), used tobacco (78% vs 70%; P <.001), and had HIV (1.3% vs 0.4%; P = .002) were more likely to be tested. Infants exposed to HCV who had a higher median of well-child visits (7 vs 6; P <.001) were more likely to be tested. After accounting for maternal and infant characteristics and health care use patterns, African American infants were less likely to undergo general testing (adjusted odds ratio 0.32; 95% confidence interval, 0.13-0.78). CONCLUSIONS: Testing occurred in <1 in 4 infants exposed to HCV and less frequently among African American infants. Public health systems need to be bolstered to ensure that infants exposed to HCV are tested for seroconversion.

Development of a System for Postmarketing Population Pharmacokinetic and Pharmacodynamic Studies Using Real-World Data From Electronic Health Records.

Postmarketing population pharmacokinetic (PK) and pharmacodynamic (PD) studies can be useful to capture patient characteristics affecting PK or PD in real-world settings. These studies require longitudinally measured dose, outcomes, and covariates in large numbers of patients; however, prospective data collection is cost-prohibitive. Electronic health records (EHRs) can be an excellent source for such data, but there are challenges, including accurate ascertainment of drug dose. We developed a standardized system to prepare datasets from EHRs for population PK/PD studies. Our system handles a variety of tasks involving data extraction from clinical text using a natural language processing algorithm, data processing, and data building. Applying this system, we performed a fentanyl population PK analysis, resulting in comparable parameter estimates to a prior study. This new system makes the EHR data extraction and preparation process more efficient and accurate and provides a powerful tool to facilitate postmarketing population PK/PD studies using information available in EHRs.

medExtractR: A targeted, customizable approach to medication extraction from electronic health records.

OBJECTIVE: We developed medExtractR, a natural language processing system to extract medication information from clinical notes. Using a targeted approach, medExtractR focuses on individual drugs to facilitate creation of medication-specific research datasets from electronic health records. MATERIALS AND METHODS: Written using the R programming language, medExtractR combines lexicon dictionaries and regular expressions to identify relevant medication entities (eg, drug name, strength, frequency). MedExtractR was developed on notes from Vanderbilt University Medical Center, using medications prescribed with varying complexity. We evaluated medExtractR and compared it with 3 existing systems: MedEx, MedXN, and CLAMP (Clinical Language Annotation, Modeling, and Processing). We also demonstrated how medExtractR can be easily tuned for better performance on an outside dataset using the MIMIC-III (Medical Information Mart for Intensive Care III) database. RESULTS: On 50 test notes per development drug and 110 test notes for an additional drug, medExtractR achieved high overall performance (F-measures >0.95), exceeding performance of the 3 existing systems across all drugs. MedExtractR achieved the highest F-measure for each individual entity, except drug name and dose amount for allopurinol. With tuning and customization, medExtractR achieved F-measures >0.90 in the MIMIC-III dataset. DISCUSSION: The medExtractR system successfully extracted entities for medications of interest. High performance in entity-level extraction provides a strong foundation for developing robust research datasets for pharmacological research. When working with new datasets, medExtractR should be tuned on a small sample of notes before being broadly applied. CONCLUSIONS: The medExtractR system achieved high performance extracting specific medications from clinical text, leading to higher-quality research datasets for drug-related studies than some existing general-purpose medication extraction tools.

Increased Incidence of Resistant Hypertension in Patients With Systemic Lupus Erythematosus: A Retrospective Cohort Study.

OBJECTIVE: To compare the risk of resistant hypertension (RHTN) in patients with systemic lupus erythematosus (SLE) and in controls without SLE, and to define factors associated with RHTN in patients with SLE. METHODS: We studied 1,044 patients with SLE and 5,241 control subjects using de-identified electronic health records from a tertiary care center. SLE was defined as ≥4 International Classification of Diseases, Ninth Revision codes for SLE and antinuclear antibody titer ≥1:160. RHTN was defined as uncontrolled blood pressure on 3 antihypertensive medications or requiring 4 or more antihypertensives to attain control. First, we compared the risk of RHTN between groups. Second, we examined the association between RHTN and all-cause mortality in patients with SLE. RESULTS: RHTN was nearly twice as prevalent in patients with SLE compared to control subjects (10.2% and 5.3%, respectively), with an incidence rate of 10.2 versus 6.1 cases per 1,000 person-years of observation (hazard ratio [HR] 1.72 [95% confidence interval 1.28-2.30]; P < 0.001, adjusted for age, sex, race, baseline end-stage renal disease [ESRD], creatinine, and calendar year). In patients with SLE, we found associations between RHTN and black race, lower renal function, hypercholesterolemia, and increased inflammatory markers. RHTN was associated with a significantly higher mortality risk (HR 2.91, P = 0.0005) after adjustment for age, sex, race, calendar year, creatinine, baseline ESRD, and number of visits. CONCLUSION: Patients with SLE have a higher risk of RHTN compared to frequency-matched controls, independent of multiple covariates. RHTN is an important comorbidity for clinicians to recognize in SLE, because it is associated with a higher risk of mortality.