Predicting inpatient pharmacy order interventions using provider action data.
Ontology highlight
ABSTRACT: Objective
The widespread deployment of electronic health records (EHRs) has introduced new sources of error and inefficiencies to the process of ordering medications in the hospital setting. Existing work identifies orders that require pharmacy intervention by comparing them to a patient's medical records. In this work, we develop a machine learning model for identifying medication orders requiring intervention using only provider behavior and other contextual features that may reflect these new sources of inefficiencies.Materials and methods
Data on providers' actions in the EHR system and pharmacy orders were collected over a 2-week period in a major metropolitan hospital system. A classification model was then built to identify orders requiring pharmacist intervention. We tune the model to the context in which it would be deployed and evaluate global and local feature importance.Results
The resultant model had an area under the receiver-operator characteristic curve of 0.91 and an area under the precision-recall curve of 0.44.Conclusions
Providers' actions can serve as useful predictors in identifying medication orders that require pharmacy intervention. Careful model tuning for the clinical context in which the model is deployed can help to create an effective tool for improving health outcomes without using sensitive patient data.
SUBMITTER: Balestra M
PROVIDER: S-EPMC8490931 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BackgroundTo develop and compare different AutoML frameworks and machine learning models to predict premature birth.MethodsThe study used a large electronic medical record database to include 715,962 participants who had the principal diagnosis code of childbirth. Three Automatic Machine Learning (AutoML) were used to construct machine learning models including tree-based models, ensembled models, and deep neural networks on the training sample (N = 536,971). The area under the curve (AUC) and training times were used to assess the performance of the prediction models, and feature importance was computed via permutation-shuffling.ResultsThe H2O AutoML framework had the highest median AUC of 0.846, followed by AutoGluon (median AUC: 0.840) and Auto-sklearn (median AUC: 0.820), and the median training time was the lowest for H2O AutoML (0.14 min), followed by AutoGluon (0.16 min) and Auto-sklearn (4.33 min). Among different types of machine learning models, the Gradient Boosting Machines (GBM) or Extreme Gradient Boosting (XGBoost), stacked ensemble, and random forrest models had better predictive performance, with median AUC scores being 0.846, 0.846, and 0.842, respectively. Important features related to preterm birth included premature rupture of membrane (PROM), incompetent cervix, occupation, and preeclampsia.ConclusionsOur study highlights the potential of machine learning models in predicting the risk of preterm birth using readily available electronic medical record data, which have significant implications for improving prenatal care and outcomes.
| S-EPMC10867966 | biostudies-literature
Project description:BackgroundPrecision medicine may offer new strategies to treat migraine, and access to existing large cohorts may be a key resource to increase statistical power. Treatment response data is not routinely collected for large cohorts; however, such information could be extracted from pharmacy databases. Using a clinical migraine sample with treatment effect data, we assessed whether treatment response can be predicted based on the number of drug purchases.MethodsA clinical cohort including 1913 migraineurs were interviewed using a semi-structured interview to retrieve treatment response data for acute and prophylactic migraine drugs. The purchase history was obtained from the Danish national pharmacy database. We assessed whether number of purchases at different thresholds could predict the specificity and sensitivity of treatment response.ResultsPurchase history of drugs was significantly associated with treatment response. For triptan treatment the specificity and sensitivity were above 80% for individuals with at least ten purchases. For prophylactic treatment (beta-blockers, angiotensin II antagonists or antiepileptic) we observed a sensitivity and specificity above 80% and 50% for individuals purchasing any prophylactic drug at least four times. In the Danish pharmacy database, 73% of the migraine patients have purchased at least ten triptans, while 55-63% have purchased at least one of the four prophylactic drugs.ConclusionPharmacy databases are a valid source for identification of treatment response. Specifically for migraine drugs, we conclude that ten purchases of triptans or four purchases of prophylactic drugs are sufficient to predict a positive treatment response. Precision medicine may be accelerated with the use of pharmacy databases.
| S-EPMC6734320 | biostudies-literature
Project description:Purpose: The aim of this study was to investigate the effect of performance transparency and individualized feedback on pharmacy technician compliance with barcode verification technology during inpatient order preparation. Methods: Following the incorporation of barcode scanning technology into the workflow of pharmacy staff, a multiphasic intervention was employed to promote its use. The intervention included verbal feedback and publically posting performance metrics to increase accountability. An interrupted time-series analysis was conducted to ascertain trends and levels in the percent of orders that were dispensed using barcode verification, before and after the study intervention. Analyses were conducted by shift and overall for pharmacy workers in a single satellite pharmacy. Results: A significant increase in percent scanned orders was observed immediately following the intervention in our analysis of all pharmacy workers (+14.4%; P = .045; 95% confidence interval [CI]: 0.35-28.4). In the analysis of each shift, statistically significant increases in percent scanned orders were observed immediately following the intervention for both the evening shift (+5.1%; P = .024; 95% CI: 0.70-9.6) and the night shift (+29.9%; P = .025; 95% CI: 3.9-55.9), but not the day shift (+2.6; P = .707; 95% CI: -11.1 to 16.2). Conclusion: Increasing transparency of individual and team performance metrics in conjunction with targeted feedback is an effective intervention to improve compliance with barcode scanning technology.
| S-EPMC6333947 | biostudies-literature
Project description:In a general inpatient population, we predicted patient-specific medication orders based on structured information in the electronic health record (EHR). Data on over three million medication orders from an academic medical center were used to train two machine-learning models: A deep learning sequence model and a logistic regression model. Both were compared with a baseline that ranked the most frequently ordered medications based on a patient's discharge hospital service and amount of time since admission. Models were trained to predict from 990 possible medications at the time of order entry. Fifty-five percent of medications ordered by physicians were ranked in the sequence model's top-10 predictions (logistic model: 49%) and 75% ranked in the top-25 (logistic model: 69%). Ninety-three percent of the sequence model's top-10 prediction sets contained at least one medication that physicians ordered within the next day. These findings demonstrate that medication orders can be predicted from information present in the EHR.
| S-EPMC7325318 | biostudies-literature
Project description:Background and objectivesDespite the dramatic increase in the provision of virtual nephrology care, only anecdotal reports of outcomes without comparators to usual care exist in the literature. This study aimed to provide objective determination of clinical noninferiority of hybrid (telenephrology plus face-to-face) versus standard (face-to-face) inpatient nephrology care.Design, setting, participants, & measurementsThis retrospective study compares objective outcomes in patients who received inpatient hybrid care versus standard nephrology care at two Mayo Clinic Health System community hospitals. Outcomes were then additionally compared with those patients receiving care at another Mayo Clinic Health System site where only standard care is available. Hospitalized adults who had nephrology consults from March 1, 2020 to February 28, 2021 were considered. Regression was used to assess 30-day mortality, length of hospitalization, readmissions, odds of being prescribed dialysis, and hospital transfers. Sensitivity analysis was performed using patients who had ≥50% of their care encounters via telenephrology. Structured surveys were used to understand the perspectives of non-nephrology hospital providers and telenephrologists.ResultsIn total, 850 patients were included. Measured outcomes that included the number of hospital transfers (odds ratio, 1.19; 95% confidence interval, 0.37 to 3.82) and 30-day readmissions (odds ratio, 0.97; 95% confidence interval, 0.84 to 1.06), among others, did not differ significantly between controls and patients in the general cohort. Telenephrologists (n=11) preferred video consults (82%) to phone for communication. More than half (64%) of telenephrologists spent less time on telenephrology compared with standard care. Non-nephrology hospital providers (n=21) were very satisfied (48%) and satisfied (29%) with telenephrology response time and felt telenephrology was as safe as standard care (67%), while providing them enough information to make patient care decisions (76%).ConclusionsOutcomes for in-hospital nephrology consults were not significantly different comparing hybrid care versus standard care. Non-nephrology hospital providers and telenephrologists had favorable opinions of telenephrology and most perceived it is as safe and effective as standard care.PodcastThis article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2022_04_11_CJN13441021.mp3.
| S-EPMC9269575 | biostudies-literature
Project description:The authors describe a pragmatic approach to the introduction of clinical decision support at the point of care, based on a decade of experience in developing and evolving Vanderbilt's inpatient "WizOrder" care provider order entry (CPOE) system. The inpatient care setting provides a unique opportunity to interject CPOE-based decision support features that restructure clinical workflows, deliver focused relevant educational materials, and influence how care is delivered to patients. From their empirical observations, the authors have developed a generic model for decision support within inpatient CPOE systems. They believe that the model's utility extends beyond Vanderbilt, because it is based on characteristics of end-user workflows and on decision support considerations that are common to a variety of inpatient settings and CPOE systems. The specific approach to implementing a given clinical decision support feature within a CPOE system should involve evaluation along three axes: what type of intervention to create (for which the authors describe 4 general categories); when to introduce the intervention into the user's workflow (for which the authors present 7 categories), and how disruptive, during use of the system, the intervention might be to end-users' workflows (for which the authors describe 6 categories). Framing decision support in this manner may help both developers and clinical end-users plan future alterations to their systems when needs for new decision support features arise.
| S-EPMC1518541 | biostudies-literature
Project description:AimsAssociations between individual medication use and falling in older individuals are well-documented. However, a comprehensive risk score that takes into account overall medication use and that can be used in daily pharmacy practice is lacking. We, therefore, aimed to determine whether pharmacy dispensing records can be used to predict falls.MethodsA retrospective cohort study was conducted using pharmacy dispensing data and self-reported falls among 3454 Dutch individuals aged ≥65 years. Two different methods were used to classify medication exposure for each person: the drug burden index (DBI) for cumulative anticholinergic and sedative medication exposure as well as exposure to fall risk-increasing drugs (FRIDs). Multinomial regression analyses, adjusted for age and sex, were conducted to investigate the association between medication exposure and falling classified as nonfalling, single falling and recurrent falling. The predictive performances of the DBI and FRIDs exposure were estimated by the polytomous discrimination index (PDI).ResultsThere were 521 single fallers (15%) and 485 recurrent fallers (14%). We found significant associations between a DBI ≥1 and single falling (adjusted odds ratio: 1.30 [95% confidence interval {CI}: 1.02-1.66]) and recurrent falling (adjusted odds ratio: 1.60 [95%CI: 1.25-2.04]). The PDI of the DBI model was 0.41 (95%CI: 0.39-0.42) and the PDI of the FRIDs model was 0.45 (95%CI: 0.43-0.47), indicating poor discrimination between fallers and nonfallers.ConclusionThe study shows significant associations between medication use and falling. However, the medication-based models were insufficient and other factors should be included to develop a risk score for pharmacy practice.
| S-EPMC9328421 | biostudies-literature
Project description:BackgroundInpatient clinical registries generally have limited ability to provide a longitudinal perspective on care beyond the acute episode. We present a method to link hospitalization records from registries with Medicare inpatient claims data, without using direct identifiers, to create a unique data source that pairs rich clinical data with long-term outcome data.Methods and resultsThe method takes advantage of the hospital clustering observed in each database by demonstrating that different combinations of indirect identifiers within hospitals yield a large proportion of unique patient records. This high level of uniqueness also allows linking without advance knowledge of the Medicare provider number of each registry hospital. We applied this method to 2 inpatient databases and were able to identify 81% of 39,178 records in a large clinical registry of patients with heart failure and 91% of 6,581 heart failure records from a hospital inpatient database. The quality of the link is high, and reasons for incomplete linkage are explored. Finally, we discuss the unique opportunities afforded by combining claims and clinical data for specific analyses.ConclusionsIn the absence of direct identifiers, it is possible to create a high-quality link between inpatient clinical registry data and Medicare claims data. The method will allow researchers to use existing data to create a linked claims-clinical database that capitalizes on the strengths of both types of data sources.
| S-EPMC2732025 | biostudies-literature
Project description:ObjectiveThe primary aim of this study was to apply a novel technology acceptance model (TAM) for virtual reality (VR) in healthcare. The secondary aim was to assess reliability of this model to evaluate factors that predict the intentions of pediatric health providers' use of VR as an anxiolytic for hospitalized pediatric patients.Materials and methodsHealthcare providers that interacted with pediatric patients participated in a VR experience available as anxiolysis for minor procedures and then completed a survey evaluating attitudes, behaviors, and technology factors that influence adoption of new technologies.ResultsReliability for all domain measurements were good, and all confirmatory factor analysis models demonstrated good fit. Usefulness, ease of use, curiosity, and enjoyment of the VR experience all strongly predict intention to use and purchase VR technologies. Age of providers, past use, and cost of technology did not influence future purchase or use, suggesting that VR technologies may be broadly adopted in the pediatric healthcare setting.DiscussionPrevious VR-TAM models in non-healthcare consumers formulated that age, past use, price willing to pay, and curiosity impacted perceived ease of use. This study established that age, past use, and cost may not influence use in healthcare. Future studies should be directed at evaluating the social influences and facilitating conditions within healthcare that play a larger influence on technology adoption.ConclusionThe VR-TAM model demonstrated validity and reliability for predicting intent to use VR in a pediatric hospital.
| S-EPMC10483581 | biostudies-literature
Project description:ObjectiveProportion of Days Covered (PDC) is a measure of medication adherence that uses prescription claims data to describe the proportion of days that the patient possessed medication. The objective of this study is to compare PDC and self-report as measures of oral contraceptive pills (OCPs) adherence and to identify individual-level predictors of adherence.Study designIn a sample of 384 OCP users, self-report was compared with PDC as measures of adherence over the past 3 months. Patient-level variables were examined for associations with adherence using multivariable logistic regression models.ResultsHigh adherence, defined as missing ≤1 pill per month, was 76%, 68% and 54% as measured by self-report, PDC and both measures, respectively. Younger women (ages 18-25 and 26-33 years) were significantly less likely to have high adherence on both measures than women in the 34-40 age group [adjusted odds ratio (OR) 0.20, 95% confidence interval (CI) 0.08-0.51 and adjusted OR 0.26, 95% CI 0.11-0.62, respectively). Other predictors of high adherence on both self-report and PDC measures included being in a relationship (adjusted OR 2.30, 95% CI 1.14-4.64, compared with unpartnered women), Protestant religion (adjusted OR 2.08, 95% CI 1.07-4.06, compared with women with no religious affiliation) and higher contraceptive self-efficacy (adjusted OR 1.63, 95% CI 1.03-2.58).ConclusionPDC derived from pharmacy claims, or a combination of PDC and self-report measures, may be an alternative to self-report alone for measuring OCP adherence.ImplicationsPDC may be a potential tool for measuring women's adherence to OCPs and should be validated in future studies.
| S-EPMC6540974 | biostudies-literature