Electronic Health Record Decision Support Research Articles

A novel electronic health record-based, machine-learning model to predict severe hypoglycemia leading to hospitalizations in older adults with diabetes: A territory-wide cohort and modeling study.

Older adults with diabetes are at high risk of severe hypoglycemia (SH). Many machine-learning (ML) models predict short-term hypoglycemia are not specific for older adults and show poor precision-recall. We aimed to develop a multidimensional, electronic health record (EHR)-based ML model to predict one-year risk of SH requiring hospitalization in older adults with diabetes. We adopted a case-control design for a retrospective territory-wide cohort of 1,456,618 records from 364,863 unique older adults (age ≥65 years) with diabetes and at least 1 Hong Kong Hospital Authority attendance from 2013 to 2018. We used 258 predictors including demographics, admissions, diagnoses, medications, and routine laboratory tests in a one-year period to predict SH events requiring hospitalization in the following 12 months. The cohort was randomly split into training, testing, and internal validation sets in a 7:2:1 ratio. Six ML algorithms were evaluated including logistic-regression, random forest, gradient boost machine, deep neural network (DNN), XGBoost, and Rulefit. We tested our model in a temporal validation cohort in the Hong Kong Diabetes Register with predictors defined in 2018 and outcome events defined in 2019. Predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC) statistics, and positive predictive value (PPV). We identified 11,128 SH events requiring hospitalization during the observation periods. The XGBoost model yielded the best performance (AUROC = 0.978 [95% CI 0.972 to 0.984]; AUPRC = 0.670 [95% CI 0.652 to 0.688]; PPV = 0.721 [95% CI 0.703 to 0.739]). This was superior to an 11-variable conventional logistic-regression model comprised of age, sex, history of SH, hypertension, blood glucose, kidney function measurements, and use of oral glucose-lowering drugs (GLDs) (AUROC = 0.906; AUPRC = 0.085; PPV = 0.468). Top impactful predictors included non-use of lipid-regulating drugs, in-patient admission, urgent emergency triage, insulin use, and history of SH. External validation in the HKDR cohort yielded AUROC of 0.856 [95% CI 0.838 to 0.873]. Main limitations of this study included limited transportability of the model and lack of geographically independent validation. Our novel-ML model demonstrated good discrimination and high precision in predicting one-year risk of SH requiring hospitalization. This may be integrated into EHR decision support systems for preemptive intervention in older adults at highest risk.

Opioid Prescribing Variation After Laparoscopic Cholecystectomy in the US Military Health System

IntroductionAfter laparoscopic cholecystectomy (LC), there is a wide variation in opioid prescription miligram morphine equivalent dose (MED) and refills across US medical institutions. Given wide variation and opioid prescription guidelines, it is essential to conduct thorough health services research across medical, surgical, and patient-level factors that can be implemented to improve system-wide prescribing practices. Therefore, this study describes discharge MED variation and opioid refill probability after emergent and nonemergent LC. Materials and methodsThis retrospective cohort study included medical record data of adult patients (N = 20,025) undergoing LC from January 2016 to June 2021 in the US Military Health System. Data visualizations and bivariate analyses examined prescription patterns across hospitals and evaluated the relationship between patient-level, care-level, and system-level factors and each outcome: discharge MED and opioid refill probability. Two generalized additive mixed models evaluated the relationship between predictors and each outcome. ResultsThere was a significant variation in opioid and nonopioid pain medication prescribing practices across hospitals. While several factors were associated with discharge MED and opioid refill probability, the strongest effects were related to time period (before versus after a June 2018 Defense Health Agency policy release) and receipt of an opioid/nonopioid combination medication. Despite decreases in MED, the MED remained almost twice the recommended dose per prior research. ConclusionsVariation by hospital suggests the need for system-level changes that target genuine practice change and opioid stewardship. Inclusion of patient-reported outcomes, electronic health record decision support tools, and academic detailing programs may support system-level improvements.

Cost of Implementing an Evidence-Based Intervention to Support Safer Use of Antipsychotics in Youth.

To estimate the cost of implementing a clinical program designed to support safer use of antipsychotics in children and adolescents (youth) age 3-17years at the time of initiating an antipsychotic medication. We calculate the costs of implementing a psychiatric consultation and navigation program for youth prescribed antipsychotic medications across 4 health systems, which included an electronic health record (EHR) decision support tool, consultation with a child and adolescent psychiatrist, and up to 6months of behavioral health care navigation, as well as telemental health for patients (n = 348). Cost data were collected for both start-up and ongoing intervention phases and are estimated over a 1-year period. Data sources included study records and time-in-motion reports, analyzed from a health system perspective. Costs included both labor and nonlabor costs (2019 US dollars). The average total start-up and ongoing costs per health system were $34,007 and $185,174, respectively. The average total cost per patient was $2,128. The highest average ongoing labor cost components were telemental health ($901 per patient), followed by child and adolescent psychiatrist consultation ($659), and the lowest cost component was primary care/behavioral health provider time to review/respond to the EHR decision support tool and case consultation ($24). For health systems considering programs to promote safer and targeted use of antipsychotics among youth, this study provides estimates of the full start-up and ongoing costs of an EHR decision support tool, psychiatric consultation service, and psychotherapeutic services for patients and families.Trial registration: Clinicaltrials.gov, NCT03448575.

Integrating Alcohol-Related Prevention and Treatment Into Primary Care

Unhealthy alcohol use is common and affects morbidity and mortality but is often neglected in medical settings, despite guidelines for both prevention and treatment. To test an implementation intervention to increase (1) population-based alcohol-related prevention with brief interventions and (2) treatment of alcohol use disorder (AUD) in primary care implemented with a broader program of behavioral health integration. The Sustained Patient-Centered Alcohol-Related Care (SPARC) trial was a stepped-wedge cluster randomized implementation trial, including 22 primary care practices in an integrated health system in Washington state. Participants consisted of all adult patients (aged ≥18 years) with primary care visits from January 2015 to July 2018. Data were analyzed from August 2018 to March 2021. The implementation intervention included 3 strategies: practice facilitation; electronic health record decision support; and performance feedback. Practices were randomly assigned launch dates, which placed them in 1 of 7 waves and defined the start of the practice's intervention period. Coprimary outcomes for prevention and AUD treatment were (1) the proportion of patients who had unhealthy alcohol use and brief intervention documented in the electronic health record (brief intervention) for prevention and (2) the proportion of patients who had newly diagnosed AUD and engaged in AUD treatment (AUD treatment engagement). Analyses compared monthly rates of primary and intermediate outcomes (eg, screening, diagnosis, treatment initiation) among all patients who visited primary care during usual care and intervention periods using mixed-effects regression. A total of 333 596 patients visited primary care (mean [SD] age, 48 [18] years; 193 583 [58%] female; 234 764 [70%] White individuals). The proportion with brief intervention was higher during SPARC intervention than usual care periods (57 vs 11 per 10 000 patients per month; P < .001). The proportion with AUD treatment engagement did not differ during intervention and usual care (1.4 vs 1.8 per 10 000 patients; P = .30). The intervention increased intermediate outcomes: screening (83.2% vs 20.8%; P < .001), new AUD diagnosis (33.8 vs 28.8 per 10 000; P = .003), and treatment initiation (7.8 vs 6.2 per 10 000; P = .04). In this stepped-wedge cluster randomized implementation trial, the SPARC intervention resulted in modest increases in prevention (brief intervention) but not AUD treatment engagement in primary care, despite important increases in screening, new diagnoses, and treatment initiation. ClinicalTrials.gov Identifier: NCT02675777.

"Help Me Sleep": A Quality Initiative to Reduce Overnight Vital Signs.

Sleep is an essential part of the recovery process, yet inpatient sleep quality is poor. Patients and families report that vital signs are the most bothersome overnight disruption. Obtaining vital signs every 4 hours (Q4H) is not evidence-based and is frequently ordered indiscriminately. We aimed to decrease the percentage of patient nights with vital sign checks between 12 am and 6 am in a low-risk population from 98% to 70% within 12 months to minimize overnight sleep disruptions and improve inpatient sleep. We conducted a quality improvement project on 3 pediatric hospital medicine teams at a large free-standing children's hospital. Our multidisciplinary team defined low-risk patients as those admitted for hyperbilirubinemia and failure to thrive. Interventions were focused around education, electronic health record decision support, and patient safety. The outcome measure was the percentage of patient nights without a vital sign measurement between 12 am and 6 am and was analyzed by using statistical process control charts. Our process measure was the use of an appropriate vital sign order. Balancing measures included adverse patient events, specifically code blues outside the ICU and emergent transfers. From March 2020 to April 2021, our pediatric hospital medicine (PHM) services admitted 449 low-risk patients for a total of 1550 inpatient nights. The percentage of patient nights with overnight vital signs decreased from 98% to 38%. There were no code blues or emergent transfers. Our improvement interventions reduced the frequency of overnight vital sign monitoring in 2 low-risk groups without any adverse events.

Using pharmacy technicians and electronic health record capabilities to improve outcomes for patients with cardiovascular disease

ObjectiveThis study aimed to compare lipid and blood pressure (BP) control before and after implementing a certified pharmacy technician (CPhT) protocol that optimized electronic health record (EHR) capabilities and shifted work from clinical pharmacy specialists (CPSs) to CPhT. SettingKaiser Permanente Colorado’s pharmacist-managed cardiac risk reduction service (which manages dyslipidemia, hypertension, and diabetes for all patients with atherosclerotic cardiovascular disease). Practice descriptionIn 2019, a protocol that optimized EHR capabilities and allowed work to be offloaded from CPS to CPhT was implemented. Filtered views within the EHR were created that bucketed patients with specific lipid results criteria. The CPhT protocol provided guidance to CPhT on determining whether patients were at low-density lipoprotein cholesterol (LDL-C) and non–high-density lipoprotein (non-HDL) goals, on appropriate statin intensity, adherent to medications, and whether the most recent BP was controlled. The CPhT notified CPS of uncontrolled patients who would assess and manage these patients, as necessary. The CPhT notified controlled patients of their results. Practice innovationData on the outcomes of incorporating pharmacy technicians to support CPS clinical activities in ambulatory clinical pharmacy are limited. Evaluation dethodsThis retrospective study compared a “Pharmacist-Driven” (index date: January 1, 2016) with a “Tech-Enhanced” (index date: January 1, 2019) group. The primary outcome was the proportion of patients at all goals defined as LDL-C < 70 mg/dL, non-HDL < 100 mg/dL, and BP < 140/90 mm Hg at 1 year after the index dates. ResultsThere were 6813 patients included (mean age: 70.2 ± 11.1 years, 71.4% male): 3130 and 3683 in the “Pharmacist-Driven” and “Tech-Enhanced” groups, respectively. The proportion of patients who attained LDL-C, non-HDL, and BP goals was higher in the “Tech-Enhanced” group (51.1% vs. 39.7%, P < 0.001) than the “Pharmacist-Driven” group. ConclusionA protocol integrating EHR decision support and CPhTs enabled work to shift to from CPS to CPhT and improved clinical outcomes.

Rationale and design of the Novel Uses of adaptive Designs to Guide provider Engagement in Electronic Health Records (NUDGE-EHR) pragmatic adaptive randomized trial: a trial protocol

BackgroundThe prescribing of high-risk medications to older adults remains extremely common and results in potentially avoidable health consequences. Efforts to reduce prescribing have had limited success, in part because they have been sub-optimally timed, poorly designed, or not provided actionable information. Electronic health record (EHR)-based tools are commonly used but have had limited application in facilitating deprescribing in older adults. The objective is to determine whether designing EHR tools using behavioral science principles reduces inappropriate prescribing and clinical outcomes in older adults.MethodsThe Novel Uses of Designs to Guide provider Engagement in Electronic Health Records (NUDGE-EHR) project uses a two-stage, 16-arm adaptive randomized pragmatic trial with a “pick-the-winner” design to identify the most effective of many potential EHR tools among primary care providers and their patients ≥ 65 years chronically using benzodiazepines, sedative hypnotic (“Z-drugs”), or anticholinergics in a large integrated delivery system. In stage 1, we randomized providers and their patients to usual care (n = 81 providers) or one of 15 EHR tools (n = 8 providers per arm) designed using behavioral principles including salience, choice architecture, or defaulting. After 6 months of follow-up, we will rank order the arms based upon their impact on the trial’s primary outcome (for both stages): reduction in inappropriate prescribing (via discontinuation or tapering). In stage 2, we will randomize (a) stage 1 usual care providers in a 1:1 ratio to one of the up to 5 most promising stage 1 interventions or continue usual care and (b) stage 1 providers in the unselected arms in a 1:1 ratio to one of the 5 most promising interventions or usual care. Secondary and tertiary outcomes include quantities of medication prescribed and utilized and clinically significant adverse outcomes.DiscussionStage 1 launched in October 2020. We plan to complete stage 2 follow-up in December 2021. These results will advance understanding about how behavioral science can optimize EHR decision support to improve prescribing and health outcomes. Adaptive trials have rarely been used in implementation science, so these findings also provide insight into how trials in this field could be more efficiently conducted.Trial registrationClinicaltrials.gov (NCT04284553, registered: February 26, 2020)

Effect of an Electronic Health Record Decision Support Alert to Decrease Excess Cervical Cancer Screening.

Cervical cancer screening is often conducted in excess of current screening guidelines. The objective of this study was to evaluate the effect of an electronic health record (EHR) clinical decision support alert to decrease guideline-nonadherent cervical cancer screening beyond the age limits of screening or posthysterectomy. The proportion of guideline-nonadherent Pap tests in women younger than 21 years or older than 65 years or posthysterectomy were compared 4 months before and 3 months after implementation of an EHR clinical decision support alert warning providers that a Pap test is not indicated. Providers could cancel the Pap test or override the alert and place the order. Provider characteristics and Pap test indications were summarized by preintervention/postintervention period using descriptive statistics. The proportions of nonindicated Pap tests were compared by intervention period and provider characteristics using generalized estimating equation models. In women beyond the screening age limits or posthysterectomy, a total of 388 Pap tests were ordered before intervention, and 313 tests were ordered after intervention. Proportion of guideline-nonadherent tests was similar before (62%) and after intervention (63%); thus, implementation of the clinical decision support alert did not change the proportion of guideline-nonadherent Pap tests ordered (OR = 1.08, 95% CI = 0.77-1.52). It is notable that 52% of guideline-nonadherent tests were ordered by 11 providers. Even when controlling for providers who ordered more than 1 test during the study period, multivariate analysis showed that male providers were more likely to order guideline-nonadherent Pap tests (OR = 2.30, 95% CI = 1.36-3.89); no other differences by provider characteristics were observed. An EHR clinical decision support alert does not decrease guideline-nonadherent cervical cancer screening. These data suggest efforts to optimize clinical decision support should be focused on other aspects of cervical cancer prevention.

Electronic Health Record-Assisted Reflex Urine Culture Testing Improves Emergency Department Diagnostic Efficiency.

The authors evaluated the effectiveness of an electronic health record (EHR)-based reflex urine culture testing algorithm on urine test utilization and diagnostic yield in the emergency department (ED). The study implemented a reflex urine culture order with EHR decision support. The primary outcome was the number of urine culture orders per 100 ED visits. The secondary outcome was the diagnostic yield of urine cultures. After the intervention, the mean number of urine cultures ordered was 5.95 fewer per 100 ED visits (9.3 vs 15.2), and there was a decrease in normal, or negative, cultures by 2.42 per 100 ED visits. There also was a statistically significant decrease in urine culture utilization and an increase in the positive proportion of cultures. Simple EHR clinical decision-support tools along with reflex urine culture testing can significantly reduce the number of urine cultures performed while improving diagnostic yield in the ED.

Integration of screening, assessment, and treatment for cannabis and other drug use disorders in primary care: An evaluation in three pilot sites

BackgroundThis pilot study evaluated whether use of evidence-based implementation strategies to integrate care for cannabis and other drug use into primary care (PC) as part of Behavioral Health Integration (BHI) increased diagnosis and treatment of substance use disorders (SUDs). MethodsPatients who visited the three pilot PC sites were eligible. Implementation strategies included practice coaching, electronic health record decision support, and performance feedback (3/2015-4/2016). BHI introduced annual screening for past-year cannabis and other drug use, a Symptom Checklist for DSM-5 SUDs, and shared decision-making about treatment options. Main analyses tested whether the proportions of PC patients diagnosed with, and treated for, new cannabis or other drug use disorders (CUDs and DUDs, respectively), differed significantly pre- and post-implementation. ResultsOf 39,599 eligible patients, 57% and 59% were screened for cannabis and other drug use, respectively. Among PC patients reporting daily cannabis use (2%) or any drug use (1%), 51% and 37%, respectively, completed an SUD Symptom Checklist. The proportion of PC patients with newly diagnosed CUD increased significantly post-implementation (5 v 17 per 10,000 patients, p < 0.0001), but not other DUDs (10 vs 13 per 10,000, p = 0.24). The proportion treated for newly diagnosed CUDs did not increase post-implementation (1 vs 1 per 10,000, p = 0.80), but did for those treated for newly diagnosed other DUDs (1 vs 3 per 10,000, p = 0.038). ConclusionsA pilot implementation of BHI to increase routine screening and assessment for SUDs was associated with increased new CUD diagnoses and a small increase in treatment of new other DUDs.

Quality Improvement in Pediatric Head Trauma with PECARN Rules Implementation as Computerized Decision Support.

For the 1.4 million emergency department (ED) visits for traumatic brain injury (TBI) annually in the United States, computed tomography (CT) may be over utilized. The Pediatric Emergency Care Applied Research Network developed 2 prediction rules to identify children at very low risk of clinically important TBI. We implemented these prediction rules as decision support within our electronic health record (EHR) to reduce CT. To test EHR decision support implementation in reducing CT rates for head trauma at 2 pediatric EDs. We compared monthly CT rates 1 year before [preimplementation (PRE)] and 1 year after [postimplementation (POST)] decision support implementation. The primary outcome was change in CT use rate over time, measured using statistical process control charts. Secondary analyses included multivariate comparisons of PRE to POST. Balancing measures included ED length of stay and returns within 7 days after ED release. There were 2,878 patients with head trauma (1,329 PRE and 1,549 POST) included. Statistical process control charts confirmed decreased CT rates over time POST that was not present PRE. Secondary statistical analyses confirmed that CT scan utilization rates decreased from 26.8% to 18.9% (unadjusted Odds Ratio [OR], 0.64; 95% Confidence Interval [CI], 0.53 -0.76; adjusted OR, 0.71; 95% CI, 0.58 -0.86). Length of stay was unchanged. There was no increase in returns within 7 days and no significant missed diagnoses. Implementation of EHR-integrated decision support for children with head trauma presenting to the ED is associated with a decrease in CT utilization and no increase in significant safety events.

Implementation of Clinical Pharmacogenomics within a Large Health System: From Electronic Health Record Decision Support to Consultation Services.

The number of clinically relevant gene-based guidelines and recommendations pertaining to drug prescribing continues to grow. Incorporating gene-drug interaction information into the drug-prescribing process can help optimize pharmacotherapy outcomes and improve patient safety. However, pharmacogenomic implementation barriers exist such as integration of pharmacogenomic results into electronic health records (EHRs), development and deployment of pharmacogenomic decision support tools to EHRs, and feasible models for establishing ambulatory pharmacogenomic clinics. We describe the development of pharmacist-managed pharmacogenomic services within a large health system. The Clinical Pharmacogenetics Implementation Consortium guidelines for HLA-B*57:01-abacavir, HLA-B*15:02-carbamazepine, and TPMT-thiopurines (i.e., azathioprine, mercaptopurine, and thioguanine) were systematically integrated into patient care. Sixty-three custom rules and alerts (20 for TPMT-thiopurines, 8 for HLA-B*57:01-abacavir, and 35 for HLA-B*15:02-anticonvulsants) were developed and deployed to the EHR for the purpose of providing point-of-care pharmacogenomic decision support. In addition, a pharmacist and physician-geneticist collaboration established a pharmacogenomics ambulatory clinic. This clinic provides genetic testing when warranted, result interpretation along with pharmacotherapy recommendations, and patient education. Our processes for developing these pharmacogenomic services and solutions for addressing implementation barriers are presented.

Inadequate Systems to Support Breast and Cervical Cancer Screening in Primary Care Practice.

Despite substantial resources devoted to cancer screening nationally, the availability of clinical practice-based systems to support screening guidelines is not known. To characterize the prevalence and correlates of practice-based systems to support breast and cervical cancer screening, with a focus on the patient-centered medical home (PCMH). Web and mail survey of primary care providers conducted in 2014. The survey assessed provider (gender, training) and facility (size, specialty training, physician report of National Committee for Quality Assurance (NCQA) PCMH recognition, and practice affiliation) characteristics. A hierarchical multivariate analysis clustered by clinical practice was conducted to evaluate characteristics associated with the adoption of practice-based systems and technology to support guideline-adherent screening. Primary care physicians in family medicine, general internal medicine, and obstetrics and gynecology, and nurse practitioners or physician assistants from four clinical care networks affiliated with PROSPR (Population-based Research Optimizing Screening through Personalized Regimens) consortium research centers. The prevalence of routine breast cancer risk assessment, electronic health record (EHR) decision support, comparative performance reports, and panel reports of patients due for routine screening and follow-up. There were 385 participants (57.6 % of eligible). Forty-seven percent (47.0 %) of providers reported NCQA recognition as a PCMH. Less than half reported EHR decision support for breast (48.8 %) or cervical cancer (46.2 %) screening. A minority received comparative performance reports for breast (26.2 %) or cervical (19.7 %) cancer screening, automated reports of patients overdue for breast (18.7 %) or cervical (16.4 %) cancer screening, or follow-up of abnormal breast (18.1 %) or cervical (17.6 %) cancer screening tests. In multivariate analysis, reported NCQA recognition as a PCMH was associated with greater use of comparative performance reports of guideline-adherent breast (OR 3.23, 95 % CI 1.58-6.61) or cervical (OR 2.56, 95 % CI 1.32-4.96) cancer screening and automated reports of patients overdue for breast (OR 2.19, 95 % CI 1.15-41.7) or cervical (OR. 2.56, 95 % CI 1.26-5.26) cancer screening. Providers lack systems to support breast and cervical cancer screening. Practice transformation toward a PCMH may support the adoption of systems to achieve guideline-adherent cancer screening in primary care settings.

Does Money Really Talk? A Resident-Driven Pay-for-Performance Pilot.

I n this issue of the Journal of Graduate Medical Education, an article entitled ‘‘Successful resident engagement in quality improvement: the Detroit Medical Center story’’ by Hussain et al presents an interesting foray into the world of payfor-performance. The authors focused on a residentdriven pay-for-performance initiative that targeted venous thromboembolism (VTE) prevention and stroke care in a large urban academic medical center. In addition to being predominantly resident-driven, the intervention is unique, in that it successfully obtained full institutional support of an educational endeavor. It thus serves as an excellent example of what can be achieved when educational and institutional goals are aligned. The article also highlights the importance of adopting adjunctive software that augments established electronic health record systems to improve the quality of care delivered to a target population. In the pilot described by the authors, a decision support tool was utilized that tailored evidence-based recommendations for VTE prevention and stroke care to the patients being evaluated. The study offers a first look into whether resident behavior can be modified by the use of financial incentives. We know from earlier research that physicians’ performance can be changed, for better or for worse, based on financial incentives. The most illustrative example of financially motivated behavioral changes comes from a 2004 United Kingdom (UK) experiment. Family medicine physicians were incentivized to adhere to 136 clinically based core measures, known collectively as the Quality and Outcomes Framework. As we now know, the results were astounding, in that the payouts reached 83.4% of available incentive payments within the first year of the program and increased to 97.8% by 2007. Hussain et al have shown that resident behavior can similarly be altered by financial motivations. As educators, it is encouraging to know that an educational program can achieve results of this magnitude across multiple disciplines and at the same time be completely self-policed and self-maintained. Not only does this exemplify the importance of institutional buy-in, but it also demonstrates the success that can be achieved through engagement of front-line staff rather than through blanket edicts. To date, most residency programs have predominately payed lip service to the quality movement, asking trainees to understand the relevance of quality metrics and pay-for-performance by using hypothetical scenarios rather than real life situations. While this program demonstrated resounding success in achieving the designated performance measures, a number of concerns spring to light related to the integrity of the pay-for-performance concept. The performance metrics used in the authors’ study were process measures, chosen for their ease of measurement and the ease with which they can define success. The reader is left with the age-old question of whether achieving these measures had an impact on clinically relevant outcomes, and if so, were the outcomes such that the financial input necessary to achieve them was reasonable. We know that to implement such a broad and far-reaching residentrun program using additional electronic health record decision support, a $250,000 startup cost was required. It would be interesting to know whether the hospital realized equivalent cost savings as a result of this initiative. Similar to questions raised by the UK experiment, with the intervention described by Hussain et al there are issues to be considered on a local level. The achievement gap discussed in this article was extremely narrow, with compliance rates for VTE performance measures increasing from 88.5% at baseline, to 94.2% at 6 months, and an astounding 100% at 12 months. Similarly, the performance measures for stroke care improved from a baseline of 88%, to 96.6% in 6 months, to 100% compliance at 12 months. While impressive, one could argue that with the preimplementation baseline being so high, such improvements have little practical meaning in terms of both cost savings and patient outcome benefits. Readers also should be wary any time a performance indicator achieves and maintains longterm sustainability at 100% compliance. While we revel in these outcomes, ‘‘to err is human,’’ even with decision support programs. The 100% success rates compel us to ask: Were there patients who were inappropriately excluded from the denominator DOI: http://dx.doi.org/10.4300/JGME-D-16-00105.1

Improving Medication Safety in Primary Care Using Electronic Health Records

Electronic health record (EHR) systems offer promising tools to assist clinicians and staff with improving medication safety, yet many of the decision support components within these information systems are not well used. The aim of this study was to identify the strategies planned by primary care practices participating in a 2-year medication safety quality improvement intervention within the Practice Partner Research Network. A theoretical model for primary care practice improvement was used to foster team-based approaches to prioritizing performance, system redesign, better use of EHR tools, and patient activation. The intervention included network meetings, site visits and performance reports. Improvement plans were qualitatively evaluated from field notes and organized to present a comprehensive approach to improving medication safety in primary care using EHRs. A total of 32 distinct plans and 11 common strategies were developed by practices to improve adherence with prescribing and monitoring indicators. Common plans included enhancing medication reconciliation to improve the accuracy of medication lists, using Practice Partner Research Network reports to identify patients meeting criteria for preventable medication errors, and customizing and applying EHR decision support tools for medication dosing, drug-disease interactions, and monitoring. Medication safety might be improved by implementing specific strategies within the primary care setting. Further evaluation is needed to provide an evidence base for improvement.

Electronic Health Record Decision Support and Quality of Care for Children With ADHD

The objective of this study was to assess the effect of electronic health record (EHR) decision support on physician management and documentation of care for children with attention-deficit/hyperactivity disorder (ADHD). This study involved 79 general pediatricians in 12 pediatric primary care practices that use the same EHR who were caring for 412 children who were aged 5 to 18 years and had a previous diagnosis of ADHD. We conducted a cluster randomized trial of EHR-based decision support that included (1) clinician reminders to assess ADHD symptoms every 3 to 6 months and (2) an ADHD note template with structured fields for symptoms, treatment effectiveness, and adverse effects. The main outcome measures were (1) proportion of children with visits during the 6-month study period in which ADHD was assessed and (2) quality of documentation of ADHD assessment. Generalized estimating equations were used to control for the clustering by providers. Children at intervention sites were more likely to have had a visit during the study period in which their ADHD was assessed. The ADHD template was used at 32% of visits at which patients were scheduled specifically for ADHD assessment, and its use was associated with improved documentation of symptoms, treatment effectiveness, and treatment adverse effects. EHR-based decision support improved the likelihood that children with ADHD had visits for as well as care related to managing this condition. Better understanding of how to optimize provider use of the decision support and templates could promote additional improvements in care.