Clinical Decision Support for Recognition and Management of Hypertension: A Randomized Trial
OBJECTIVES: Although blood pressure (BP) is routinely measured in outpatient visits, elevated BP and hypertension are often not recognized. We evaluated whether an electronic health record–linked clinical decision support (CDS) tool could improve the recognition and management of hypertension in adolescents.
METHODS: We randomly assigned 20 primary care clinics within an integrated care system to CDS or usual care. At intervention sites, the CDS displayed BPs and percentiles, identified incident hypertension on the basis of current or previous BPs, and offered tailored order sets. The recognition of hypertension was identified by an automated review of diagnoses and problem lists and a manual review of clinical notes, antihypertensive medication prescriptions, and diagnostic testing. Generalized linear mixed models were used to test the effect of the intervention.
RESULTS: Among 31 579 patients 10 to 17 years old with a clinic visit over a 2-year period, 522 (1.7%) had incident hypertension. Within 6 months of meeting criteria, providers recognized hypertension in 54.9% of patients in CDS clinics and 21.3% of patients in usual care (P ≤ .001). Clinical recognition was most often achieved through visit diagnoses or documentation in the clinical note. Within 6 months of developing incident hypertension, 17.1% of CDS subjects were referred to dieticians or weight loss or exercise programs, and 9.4% had additional hypertension workup versus 3.9% and 4.2%, respectively (P = .001 and .046, respectively). Only 1% of patients were prescribed an antihypertensive medication within 6 months of developing hypertension.
CONCLUSIONS: The CDS had a significant, beneficial effect on the recognition of hypertension, with a moderate increase in guideline-adherent management.
- BP —
- blood pressure
- BPA —
- best-practice advisory
- CDS —
- clinical decision support
- DBP —
- diastolic blood pressure
- EHR —
- electronic health record
- ICD-9 —
- International Classification of Diseases, Ninth Revision
- ICD-10 —
- International Classification of Diseases, 10th Revision
- SBP —
- systolic blood pressure
- UC —
- usual care
What’s Known on This Subject:
Hypertension is underrecognized in children and adolescents. Clinical decision support is a tool that can be used to leverage existing data in the electronic health record to promote hypertension recognition.
What This Study Adds:
In this 20-site, cluster-randomized trial, we demonstrated how an innovative, electronic health record–linked, clinical decision support tool can increase hypertension recognition and promote guideline-adherent clinical care.
Hypertension, which is defined as having blood pressure (BP) ≥95th percentile at 3 separate clinical encounters,1–3 occurs in 1% to 3% of children and adolescents.4–6 Although BP thresholds for hypertension were modified with the 2017 Clinical Practice Guideline for Screening and Management of High Blood Pressure in Children and Adolescents,3 early recognition of hypertension remains a priority.
BP is routinely measured at outpatient visits,7 yet hypertension is often not recognized.8,9 Barriers to recognition include time pressures or competing demands at clinical encounters, perceived complexities of the current BP tables, and the need to review previous BP measurements to diagnose hypertension.
Among patients with multiple visits, electronic health records (EHRs) should contain sufficient information to diagnose hypertension. However, even when EHRs are configured to display BP percentiles, information on the patterns of BP percentiles over time, previous diagnoses, and medications is not presented in a format that is useful for clinicians. As previously described, our team developed and piloted TeenBP, an EHR-linked clinical decision support (CDS) tool that significantly increased the recognition of a single BP ≥95th percentile.10 Our goals in the current study were to evaluate the effectiveness of TeenBP in improving clinical recognition and the early management of children and adolescents with incident hypertension.
We conducted a 2-year, cluster-randomized trial in 20 primary care clinics within a large, Midwestern health system. All clinics use a common EHR (EpiCare, Verona, WI) and follow standardized procedures. Heights were measured by using stadiometers and weights with standardized electronic scales. BP was routinely measured by registered nurses or medical assistants at preventive health visits and a majority of acute or follow-up visits7 by using an automated device (Omron Healthcare), with patients rested for 3 to 5 minutes before BP measurement. Staff were trained in patient positioning, cuff size, and placement of the BP cuff. Repeat measurement of BP was recommended if the initial BP was ≥95th percentile. Across sites, when a height and BP were recorded at the same visit, BP percentiles were calculated and displayed within the EHR; BPs ≥95th percentile were displayed in red with an exclamation mark.
Patients between the ages of 10 and 17 years who met clinical criteria for incident hypertension during primary care visits at study sites between April 15, 2014 and April 14, 2016 were eligible for inclusion. Exact BP percentiles were calculated consistent with the Fourth Report1 and the Integrated Guidelines for Cardiovascular Health and Risk Reduction in Children and Adolescents, the guidelines that were in use during the time of the study.2 For each outpatient visit during the study, if the systolic blood pressure (SBP) or diastolic blood pressure (DBP) was ≥95th percentile when using automated methods, we then searched for BPs at primary care visits in the previous 2 years along with previous diagnoses of hypertension (International Classification of Diseases, Ninth Revision [ICD-9] 401-405 and International Classification of Diseases, 10th Revision [ICD-10] I10). Patients who were pregnant or ≤12 weeks postpartum were excluded. Patients with ≥3 BPs ≥95th percentile within a 2-year window and with no previous hypertension diagnoses were classified as having incident hypertension.
The random assignment of clinics took place 4 weeks before the implementation of the CDS. We first stratified sites into 3 groups on the basis of the proportion of patients with public insurance. We then sorted clinics into pairs by adolescent patient volume. By using a random-number generator, 1 clinic within each pair was randomly selected to receive the TeenBP CDS while the other continued with usual care (UC), resulting in 10 clinics assigned to the TeenBP CDS and 10 clinics assigned to UC.
As previously described,10,11 the CDS was developed over a 12-month period and piloted at a single clinic within the same health system starting in November 2013. The timing of the best-practice advisories (BPAs), wording and visual display of the CDS, and recommendations were developed to be consistent with the BP guidelines2 available at the time of the study while accounting for clinic workflow and the preferences of medical group leaders. Modifications were made in an iterative process during the pilot period. The full intervention went live at the 10 intervention sites on April 15, 2014.
The study principal investigator (E.O.K.) and project manager (H.L.E.) conducted in-person training at the 10 CDS intervention sites within 4 weeks of the start of the intervention. Rooming staff, registered nurses, and providers were encouraged to attend the 1-hour sessions. On a monthly basis, feedback was provided to nurse managers regarding their clinics’ (1) rate of repeating initial elevated-BP measures at clinic visits and (2) proportion of targeted visits at which medical providers opened the CDS when prompted. To account for staff turnover, we conducted a second round of on-site trainings at the 10 intervention sites during the first or second quarter of the second study year.
The TeenBP CDS included up to 3 BPAs for the rooming staff. If height was not documented within the previous 12 months, a height measurement was requested. Second, if the SBP or DBP at a visit was ≥95th percentile, a second BP measurement was requested. Third, if 2 measurements were obtained at a visit, a BPA would notify the rooming staff that either (1) the average of the initial and repeat BP is not elevated, so no further action is needed, or (2) the average of the initial and repeat BP is elevated and please notify the provider. When the average of ≥2 BPs at a single visit was ≥95th percentile on opening the chart, the provider received a single BPA with links to the full TeenBP CDS and a patient-tailored order set with suggested diagnoses and orders (Fig 1). When a BP ≥95th percentile was not repeated at a visit, the CDS was not delivered to the provider.
For patients with incident hypertension, the CDS recommendations included the following: a diagnosis of hypertension, lipid screening (if none were recorded in previous year), and a nutrition referral. In patients with BPs ≥99th percentile and 5 mm Hg (stage 2 hypertension) or BMI <85th percentile, the CDS advised that a workup for secondary hypertension should be considered, but specific orders were not included.
The earliest date within the study period when a patient met criteria for incident hypertension was assigned as the index date. Outcomes were assessed over 6 months after the index date. The primary outcome was the recognition of hypertension. This was a composite outcome based on at least 1 of the following: (1) outpatient discharge diagnosis of hypertension (ICD-9 401-405 and ICD-10 I10) or elevated BP (ICD-9 796.2 and ICD-10 R03.0), (2) hypertension or elevated BP documented in the clinical note, (3) hypertension or elevated BP documented in the after-visit summary, (4) hypertension or elevated BP added to the problem list, (5) a workup for secondary causes of hypertension or end-organ damage, or (6) antihypertensive medication being prescribed.
Clinician documentation was assessed through a chart review conducted by trained nurse chart abstractors blinded to study arm. Outpatient diagnoses and changes to the problem list were assessed through automated databases. Additional diagnostic workups and prescriptions were assessed through a 2-step process. First, orders for echocardiograms, renal ultrasounds, and antihypertensive medications were assessed through a chart review by trained nurse chart abstractors. Second, all diagnostic testing and prescriptions were evaluated by the study principal investigator (E.O.K.), and complicated cases were reviewed with other clinicians on the study team (P.L.F., P.O’C., and (J.D.N.) to ensure that hypertension or elevated BP was the indication for the test or prescription. Referrals to dietitians, weight loss programs, or exercise programs were assessed through a chart review. Among patients with incident hypertension and no lipid screening in the previous year or before dyslipidemia diagnosis, an order for lipid screening was also a diagnostic outcome.
Provider satisfaction with the TeenBP CDS was assessed through a brief paper survey. A $50 gift card was included with each survey. The survey included questions related to the frequency of best-practice alerts, usefulness of individual features of the CDS, and overall satisfaction with the CDS.
On the basis of a priori estimates, including a baseline hypertension recognition rate of 18%, this study was powered at 80% (2-sided α = .05) to detect a 23% absolute increase in hypertension recognition in CDS clinics. The sample size was based on an expected 163 children and adolescents with incident hypertension from 18 clinics and an expected intraclass correlation of 0.05 for hypertension recognition. Differences by study arm in baseline characteristics were tested with Pearson’s χ2 test, Fisher’s exact tests, and independent sample t tests. Generalized linear mixed models with a logit link and a random intercept for clinic were used to test the effect of the intervention on binary outcomes of the clinical recognition of hypertension. Each model included a binary indicator for study arm and the following a priori–selected covariates: SBP percentile at the index visit, BMI percentile at the index visit, and index visit type (well-child or acute and/or follow-up visit). Adjusted, model-predicted proportions and 95% confidence intervals are reported for each outcome. Similar models were developed to evaluate differences in diagnostic workups, referrals, and lipid screenings by study arm. All analyses were 2 sided, and P values of <.05 were considered statistically significant. Corrections for multiple comparisons were not conducted. All analyses were conducted with SAS version 9.4 (SAS Institute, Inc, Cary, NC). Survey results were evaluated by using descriptive statistics.
This study was approved by the HealthPartners Institute Institutional Review Board with a waiver of informed consent for the implementation and evaluation of the CDS. Provider consent was implied through their return of the survey.
Among the 10 TeenBP CDS clinics and between April 15, 2014 and April 14, 2016, there were 17 037 patients 10 to 17 years of age with at least 1 BP recorded. Of these, 296 (1.7%) had incident hypertension. At the 10 UC clinics over the same time period, there were 14 542 patients 10 to 17 years of age with at least 1 BP recorded and 226 (1.6%) with incident hypertension (Fig 2). Of the 31 579 patients with at least 1 BP, the majority (63.5%) were white, non-Hispanic (94.2%), and privately insured (92.6%). The final population meeting the criteria for incident hypertension was similar to the larger cohort with the exception of having a higher proportion of girls (56.7% vs 49.7%). Patient and visit characteristics did not vary significantly between the CDS and UC groups (Table 1).
Within 6 months of meeting the criteria for incident hypertension, 54.9% of the CDS patients versus 21.3% of the UC patients were clinically recognized (P < .001). Recognition was most commonly achieved by having hypertension or elevated BP as a discharge diagnosis at an outpatient encounter, in a clinical note, or in the patient discharge instructions (Table 2). Only 10.1% of patients at intervention clinics and 5.2% of patients at UC sites had hypertension or elevated BP added to their problem lists.
Evaluations for secondary causes of hypertension or target organ damage were uncommon overall but were more common in intervention versus UC populations (9.4% vs 4.2%; P = .046). Referrals to dietitians or weight loss or exercise programs were more common in CDS patients (17.1%) versus UC patients (3.9%; P = .001). Among the 256 CDS patients and 205 UC patients who were eligible for lipid screening, 14.4% of CDS versus 5.3% of UC patients had lipid screenings completed (P = .03). The provision of antihypertensive medication within 6 months of meeting the criteria occurred in 1% of intervention patients and 0.4% of UC patients (P = .48).
Among 86 primary care providers at CDS intervention clinics, 71 (83%) completed a brief survey between March 2016 and June 2016. Eighty percent of respondents were physicians, and 20% were advanced practice providers; 65% had >10 years of experience in clinical practice. A majority of the respondents (75%) recalled interacting with the CDS. Of those who reported using the CDS, 92% thought it was useful in identifying patients with elevated BP or hypertension, 94% agreed that time using the CDS was “time well spent,” and 95% agreed that the CDS tool was useful for shared decision-making.
In this 2-year, cluster-randomized, pragmatic trial, an innovative, EHR-linked CDS tool was well accepted among providers and resulted in an improved recognition of hypertension. The CDS also increased guideline-adherent care, including referrals to dieticians and cholesterol screening, but these occurred at suboptimal rates in both the intervention and UC populations.
Despite longstanding recommendations regarding the identification of hypertension in children and adolescents1,2 along with numerous studies demonstrating underdiagnosis,6,8,9,12 effective interventions to increase the recognition of pediatric hypertension have been limited. In a cohort study of visits to a single urban, academic health system, only 26% of potential case patients were diagnosed with hypertension or elevated BP.8 More recently, by using data from patients receiving care in 196 ambulatory clinics, of 3.3% of patients 3 to 18 years old meeting the criteria for hypertension, the recognition rate was 23%.6 Similarly, in the UC clinics in this study, among 226 patients with incident hypertension, only 21.3% were clinically recognized.
The implementation of the TeenBP EHR-linked CDS across the 10 CDS intervention clinics increased hypertension recognition to 54.9%. Published reports on interventions to increase the recognition of isolated, elevated BP have had a range of efficacy. A recently published study found that the recognition of elevated BP increased from 4.9% to 7.1% after the implementation of an EHR-linked BP application.13 In contrast, a real-time EHR alert increased the recognition of elevated BP at a single urban, academic center from 12.5% to 42%.14 In the latter study, improved outcomes were likely because of the use of a hard stop, forcing providers to acknowledge the alert. Given that most pediatric primary-care visits are not specifically for BP measurement and on the basis of the preferences of leaders within our health system, a less intrusive approach was chosen for this study, and substantial increases in recognition were achieved.
The TeenBP CDS was designed to integrate with clinical workflow, which is consistent with the CDS 5 Rights framework.15 As previously described,10 the CDS tool was based on the guidelines for BP management in children and adolescents1,2 in effect at the time of the study with local input from clinical and operational leaders within the medical group, and thus it contained the so-called right information. The multicomponent CDS tool included alerts for rooming staff; an alert, a tailored order set, and a link to the full CDS display for providers; and educational material for patients and their parents. This allowed for >1 right format to convey information to providers and families. Before implementation, we conducted extensive workflow assessments to ensure the various CDS components would be delivered to the right person, through the right channel, and at the right time in the workflow.
Our CDS was innovative in its display of current and previous BPs along with medications and diagnoses that may affect BP, pooling clinically useful, patient-specific data on a single screen. This information could also be printed and shared with families. Indeed, a majority of providers using the CDS reported that it was useful for shared decision-making. An additional strength is that as a Web-based tool, it will be easy to modify algorithms as guidelines or definitions for hypertension change. Furthermore, because our intervention required minimal ongoing maintenance once it was implemented, there is potential for scalability to additional clinical sites. Finally, we were aware of the potential for overtreatment; our CDS aimed to promote lifestyle interventions as the initial treatment of patients with incident hypertension. We were reassured by our case review that few initiated antihypertensive medication or underwent extensive testing over a 6-month observation period.
The TeenBP CDS was generally effective in promoting guideline-adherent care for children and adolescents with incident hypertension, with statistically significant increases in dietician referrals and cholesterol screening. Nevertheless, there is clearly a need for further improvement. Even in the intervention clinics, approximately half of the new hypertension cases were not recognized. As we update our CDS tool to incorporate the 2017 hypertension guidelines, we are also planning strategies to increase adoption of and adherence to recommendations, such as a manual trigger to access the CDS at any time during the clinical encounter. Additional work is also needed to optimize lifestyle interventions for this population. Also, consistent with pediatric hypertension guidelines, few patients were treated with antihypertensive medication within 6 months of meeting the clinical criteria for hypertension. It is possible that with a longer observation period, and similar to previous observational studies,6,16,17 an increasing number of patients would be prescribed antihypertensive medication.
Limitations of this work should be noted. First, by design, provider alerts were only delivered when ≥2 BPs were recorded at a visit and the average was ≥95th percentile. Because a repeated BP is usually lower than the initial measurement, our goal was to encourage remeasurement and increase the specificity for diagnosing of hypertension. Second, this study took place in a single health system and was tailored to the clinical workflow and culture of that system. Adaptations would be needed for use in other systems. In addition, audit feedback to encourage nurses to repeat elevated BPs and providers to open the CDS, along with 2 in-person training sessions, were conducted at intervention sites. Thus, the differences between the CDS and UC sites reflect the combined effectiveness of the CDS and the audit feedback and training.
Third, we included BPs measured at both preventive care and sick visits. Previous data from our health system found that adolescents averaged 1.6 visits per year, and one-third had no preventive health visits between the ages of 13 and 17 years.18 If BPs were limited to preventive care visits, not all patients with BPs ≥95th percentile would be seen, and the hypertension diagnoses would be based on BPs spanning 3 years or longer. It is acknowledged that by including BPs from sick visits, some patients with incident hypertension may have had acute reasons for their BP elevation.
Finally, our CDS was developed to be consistent with the available BP guidelines at the time. Some children and adolescents who were identified as hypertensive on the basis of having BPs ≥95th percentile at 3 separate visits may not meet updated definitions for hypertension, requiring BPs of ≥130/80 mm Hg.3 However, as a Web-based tool, the algorithms can be easily updated to accommodate these new clinical criteria for hypertension.
The TeenBP CDS tool, which was developed and implemented in this study, significantly increased the recognition of incident hypertension and was accepted by the providers using it. Nevertheless, further work is needed to continue to increase hypertension recognition and promote lifestyle modification in this population, which is at risk for long-term cardiovascular morbidity.
- Accepted November 1, 2017.
- Address correspondence to Elyse O. Kharbanda, MD, MPH, HealthPartners Institute, 8170 33rd Ave South, Mail Stop 23301A, Minneapolis, MN 55425. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by the National Institutes of Health (R01 HL115082 [Kharbanda]). The sponsor was not involved in the study design, data collection, analysis or interpretation of the data, writing of the article, or decision to submit for publication. Funded by the National Institutes of Health (NIH).
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2017-3756.
- National High Blood Pressure Education Program Working Group on High Blood Pressure in Children and Adolescents
- Expert Panel on Integrated Guidelines for Cardiovascular Health and Risk Reduction in Children and Adolescents
- National Heart, Lung, and Blood Institute
- Flynn JT,
- Kaelber DC,
- Baker-Smith CM, et al; Subcommittee on Screening and Management of High Blood Pressure in Children
- Cheung EL,
- Bell CS,
- Samuel JP,
- Poffenbarger T,
- Redwine KM,
- Samuels JA
- Kaelber DC,
- Liu W,
- Ross M, et al; Comparative Effectiveness Research Through Collaborative Electronic Reporting Consortium
- Beacher DR,
- Chang SZ,
- Rosen JS, et al
- Osheroff JA,
- Teich JA,
- Levick D, et al
- Nordin JD,
- Solberg LI,
- Parker ED
- Copyright © 2018 by the American Academy of Pediatrics