[Skip to Navigation]
Sign In
Invited Commentary
Health Informatics
February 4, 2022

Anticipating and Addressing Challenges During Implementation of Clinical Decision Support Systems

Author Affiliations
  • 1Division of Cardiology, Department of Medicine, Brown University Alpert Medical School, Providence, Rhode Island
JAMA Netw Open. 2022;5(2):e2146528. doi:10.1001/jamanetworkopen.2021.46528

In JAMA Network Open, Gold et al1 share results from their cluster randomized clinical trial assessing the impact of CV Wizard, an electronic health record (EHR)–embedded clinical decision support system (CDSS), on cardiovascular risk reduction in patients served by community health centers (CHCs). By implementing a previously validated CDSS in underresourced health care facilities that predominantly serve socioeconomically vulnerable patients, the authors generated 2 particularly important findings. First, they found CV Wizard was used in only 19.8% of all eligible clinical encounters. Second, they found a modest (<1%) reduction in overall 10-year atherosclerotic cardiovascular disease (ASCVD) risk with CV Wizard use only among patients with baseline ASCVD risk of greater than 20%. In this commentary, we explain why both findings highlight the importance of using established implementation science conceptual frameworks when implementing CDSS in any health care setting, particularly in underresourced health care facilities such as CHCs.

Gold et al1 implemented CV Wizard in multiple CHC organizations as an intervention to change clinician and patient behavior important for reducing patients’ ASCVD risk. In doing so, they had all the components of the practical, robust implementation and sustainability model (PRISM), a well-established implementation science framework, of Feldstein and Glasgow.2 They had an intervention (CV Wizard), recipients (CHC clinicians and patients), an external environment (CHCs), an implementation infrastructure (EHR-embedded software), and concrete measures of reach and effectiveness (rates of use and reduction in patients’ 10-year ASCVD risk). Under PRISM, successful adoption, implementation, and maintenance of any intervention depends on understanding and addressing contextual determinants (barriers and facilitators) to implementation: (1) patient and organization perspectives of the intervention, (2) relevant characteristics of the recipients (ie, CHC clinicians and patients), and (3) facilitating or barrier factors within the external environment. Before implementation of CV Wizard, Gold et al1 met with CHC clinicians and EHR programmers every 2 weeks for a full year to identify and address a wide variety of legal, compliance, technical, and clinical implementation challenges.3

Despite their extensive preimplementation efforts, Gold et al1 found that CV Wizard was used at only 19.8% of eligible clinical encounters within CHCs. Given that the intervention was an automated EHR alert containing the CV Wizard link provided to staff rooming eligible patients, and that simply viewing the information was counted as CDSS use, this means that 80% of alerts were left unheeded. Why this occurred in CHCs but not in other better-resourced health care facilities where CV Wizard rates of use have been closer to 70% to 80%4 is unclear but critical to understand. The authors state that heterogeneity in rooming protocols between CHCs impeded efforts to train rooming staff and likely contributed to decreased use of CV Wizard. Although we agree with that hypothesis, implementation science frameworks such as PRISM can help delineate alternative possibilities. For example, CHC rooming staff (recipients) may be more time-constrained (external environment) and/or more susceptible to EHR alert fatigue (staff perspective) than their counterparts in better-resourced health care facilities and thus might not have been ideal recipients of the intervention. Using this type of conceptual framework–directed thinking can prompt structured interviews of recipients to identify facilitators and barriers to implementation of any health care intervention, including CDSS, more efficiently and effectively.

The second important finding from Gold et al1 was a modest reduction in 10-year ASCVD risk with CV Wizard use only in patients with baseline risk of greater than 20%; patients with lower baseline risk did not benefit from the intervention. This finding—functionally an effectiveness metric in PRISM—can now be used by Gold et al1 to modify their adoption, implementation, and maintenance protocols. This bidirectionality is why PRISM is useful for both initial implementation and subsequent quality improvement cycles. Specifically, reach and effectiveness metrics can be used to better understand recipients and their perspectives, as well as to potentially modify the intervention itself in later iterations. For example, Gold et al1 could consider restricting CV Wizard–driven automated EHR alerts to appear only for frontline clinicians (instead of rooming staff) examining patients with a baseline 10-year ASCVD risk of greater than 20%. Focusing the intervention in this manner could not only drive higher rates of use and better effectiveness, but it could also improve the organizational perspective of the intervention by better accounting for the underresourced external environment.

In summary, Gold et al1 have provided us with valuable insights into the challenges that can be expected when attempting to implement CDSS in underresourced health care facilities such as CHCs. We believe these challenges can be anticipated—and overcome—when initial implementation and subsequent quality improvement cycles are guided by established implementation science conceptual frameworks such as PRISM.

Back to top
Article Information

Published: February 4, 2022. doi:10.1001/jamanetworkopen.2021.46528

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2022 Shah NR et al. JAMA Network Open.

Corresponding Author: Nishant R. Shah, MD, MPH, Division of Cardiology, Department of Medicine, Brown University Alpert Medical School, 830 Chalkstone Ave, Providence, RI 02908 (nishant_shah@brown.edu).

Conflict of Interest Disclosures: None reported.

Gold  R, Larson  AE, Sperl-Hillen  JM,  et al.  Effect of clinical decision support at community health centers on the risk of cardiovascular disease: a cluster randomized clinical trial.   JAMA Netw Open. 2022;5(2):e2146519. doi:10.1001/jamanetworkopen.2021.46519Google Scholar
Feldstein  AC, Glasgow  RE.  A practical, robust implementation and sustainability model (PRISM) for integrating research findings into practice.   Jt Comm J Qual Patient Saf. 2008;34(4):228-243. doi:10.1016/S1553-7250(08)34030-6 PubMedGoogle Scholar
Gold  R, Middendorf  M, Heintzman  J,  et al.  Challenges involved in establishing a web-based clinical decision support tool in community health centers.   Healthc (Amst). 2020;8(4):100488. doi:10.1016/j.hjdsi.2020.100488 PubMedGoogle Scholar
Sperl-Hillen  JM, Crain  AL, Margolis  KL,  et al.  Clinical decision support directed to primary care patients and providers reduces cardiovascular risk: a randomized trial.   J Am Med Inform Assoc. 2018;25(9):1137-1146. doi:10.1093/jamia/ocy085 PubMedGoogle ScholarCrossref