- Study protocol
- Open Access
- Open Peer Review
Rationale, design, and implementation protocol of an electronic health record integrated clinical prediction rule (iCPR) randomized trial in primary care
© Mann et al; licensee BioMed Central Ltd. 2011
- Received: 13 July 2011
- Accepted: 19 September 2011
- Published: 19 September 2011
Clinical prediction rules (CPRs) represent well-validated but underutilized evidence-based medicine tools at the point-of-care. To date, an inability to integrate these rules into an electronic health record (EHR) has been a major limitation and we are not aware of a study demonstrating the use of CPR's in an ambulatory EHR setting. The integrated clinical prediction rule (iCPR) trial integrates two CPR's in an EHR and assesses both the usability and the effect on evidence-based practice in the primary care setting.
A multi-disciplinary design team was assembled to develop a prototype iCPR for validated streptococcal pharyngitis and bacterial pneumonia CPRs. The iCPR tool was built as an active Clinical Decision Support (CDS) tool that can be triggered by user action during typical workflow. Using the EHR CDS toolkit, the iCPR risk score calculator was linked to tailored ordered sets, documentation, and patient instructions. The team subsequently conducted two levels of 'real world' usability testing with eight providers per group. Usability data were used to refine and create a production tool. Participating primary care providers (n = 149) were randomized and intervention providers were trained in the use of the new iCPR tool. Rates of iCPR tool triggering in the intervention and control (simulated) groups are monitored and subsequent use of the various components of the iCPR tool among intervention encounters is also tracked. The primary outcome is the difference in antibiotic prescribing rates (strep and pneumonia iCPR's encounters) and chest x-rays (pneumonia iCPR only) between intervention and control providers.
Using iterative usability testing and development paired with provider training, the iCPR CDS tool leverages user-centered design principles to overcome pervasive underutilization of EBM and support evidence-based practice at the point-of-care. The ongoing trial will determine if this collaborative process will lead to higher rates of utilization and EBM guided use of antibiotics and chest x-ray's in primary care.
ClinicalTrials.gov Identifier NCT01386047
- Electronic Health Record
- Chief Complaint
- Usability Testing
- Clinical Decision Support
- Electronic Health Record System
The benefits of evidence-based medicine (EBM) on the quality of clinical care and improved patient outcomes have not achieved their potential . While numerous EBM guidelines based on high-quality research have been generated and disseminated, data on their uptake into daily clinical practice have often been disappointing due to the challenges of integrating EBM recommendations into the point-of-care . As a result, EBM guidelines often end up as either cluttered paper on the wall of the medical office or idiosyncratic teaching points rarely altering clinical practice. Finding strategies to implement EBM at the point-of-care is critical as monitoring agencies and payers are increasingly using EBM guidelines as markers of quality care.
Clinical prediction rules (CPRs) are a type of EBM that uses validated rules for simple sign or symptom-based probability scores to risk stratify patients for specific prognoses and/or diagnostic assessments . While many high-quality CPRs exist, they have not been regularly implemented for day-to-day care due to inaccessibility at the point-of-care, a problem even more pronounced in the age of electronic health records (EHRs). Our search of the literature found no evidence of attempts to integrate CPRs into EHRs in the ambulatory setting, and only one instance of proposed integration in the inpatient setting . Two well-validated CPRs are the streptococcal pharyngitis (strep throat) and bacterial pneumonia CPRs [5–7]. The strep throat CPR uses five criteria (fever, swollen lymph nodes, tonsillar exudates, strep exposure, and recent cough) to estimate the probability of strep throat in a patient with a sore throat . The pneumonia criteria uses five criteria (fever, tachycardia, crackles, decreased breath sounds, and absence of asthma) to estimate the likelihood of a bacterial pneumonia in the setting of a cough . While both rules have been well-validated in the literature, their use at the point-of-care is suboptimal and new methods for incorporating them into the point-of-care are needed.
Clinical decision support (CDS) systems have been developed as platforms within EHRs to provide evidence at the point-of-care and change physician behavior . In theory, CDS should seamlessly integrate EBM into EHR systems to support the physician in delivering efficient, effective care at the point-of-care, but surprisingly has had equivocal results in ambulatory care [9–13]. Prior attempts at integrating these EBM delivery platforms into EHRs may have been limited by the lack of usability testing of the CDS interface and inadequate provider training prior to use . The lack of usability testing (i.e., useable and usefulness testing) limits the ability to assess if CDS can be effectively integrated into clinical workflow (usable) or is something desired by the clinician (usefulness). This often forces clinicians to either alter their workflow or work around the CDS tool. The lack of provider training in assessing the usability and usefulness of CDS tools and therefore how to best incorporate these tools into workflow has also limited their penetration into clinical practice. As a platform for building EBM into EHRs, CDS could significantly improve clinical workflow and quality delivery by providing access to many well-validated frontline decision aids like CPRs that are currently underutilized.
We have developed an integrated clinical prediction rules (iCPR) clinical decision support program that incorporates two well-validated CPRs (Walsh CPR for Streptococcal Pharyngitis and the Heckerling CPR for Pneumonia) into an outpatient EHR system used by the providers of nearly 40% of the nation's patients. This article discusses the design, development, usability testing, training, and implementation of study.
The iCPR study was designed to test the feasibility and effectiveness of incorporating the strep throat and pneumonia CPRs into the EHR in a primary care practice. The two main aims supporting this goal were to assess adoption of the iCPR program in primary care and to assess the impact of the iCPR
We considered several options within the EHR to house the iCPR assessment tool based on discussions with the vendor, provider familiarity with the vendor's CDS tools, and provider workflow. The EHR vendor initially suggested using a 'smart' form for iCPR because it has enhanced visual aesthetics and expedites calculations. However, providers had almost no daily experience with this form in the practice under study and, more importantly, would have required more manual input by providers to complete the full iCPR workflow. As a result, the team selected to use dynamic flowsheets for calculations that were relatively unknown and had some formatting limitations, but minimized 'clicks' and manual data entry.
Restriction of alerts
iCPR is a practice-based randomized clinical trial that had to be seamlessly integrated into workflow without disrupting control providers. To achieve this, iCPR was designed to activate only for providers randomized to the intervention. Furthermore, the tool is further restricted to the providers' outpatient primary care EHR interface, because they may be practicing in other clinical settings with the same EHR but potentially vastly different workflows.
Alerts, overrides and triggers
Alerts are an active research area in the CDS literature. They can be categorized on two spectrums of activity: active versus passive and mandatory versus optional. A major early design consideration was the whether to use active (interrupting) or passive (non-interrupting) alerts [15, 16]. In the context of our commercial EHR, active alerts 'pop-up' at the user, directly interrupting their workflow in order to draw their attention. Passive alerts are non-interrupting, minimally intrusive alerts, and would use non-interrupting flags or highlights to draw the CDS alert. Mandatory alerts require the user to take the designated action or explain the reason for overriding the CDS, while optional alerts allow the user to ignore the CDS alert without an explanation. Prior literature has demonstrated the superior efficacy of active mandatory alerts; however they are more disruptive to workflow, which contributes to the low uptake of CDS tools . Their use is also problematic in the increasingly crowded CDS dashboards populating the primary care EHR. Balancing these factors, the development team selected a two-step system in which an early-in-workflow passive mandatory alert and a later-in-workflow active mandatory alert were combined. Mandatory alerts were chosen for both because data on reasons for declining the CDS tool were critical to iterative improvements.
Chief complaint, diagnosis, and diagnosis/antibiotic combination triggers of iCPR tools
Chest hurts when breathing
Productive cough with shortness of breath
New onset shortness of breath
Recent contact (children) with pharyngitis
Acute bronchitis with bronchospasm
Difficulty in swallowing
Bronchiectasis with acute exacerbation
Pain on swallowing
Bronchitis with chronic airway obstruction
Pharyngitis due to group A beta hemolytic Streptococci
Bronchitis, not specified as acute or chronic
CAP (community acquired pneumonia)
Sore throat (viral)
Community acquired pneumonia
Sore throat - chronic
LRTI (lower respiratory tract infection)
Strep sore throat
Pneumonia, community acquired
Streptococcal sore throat
Pneumonia, organism unspecified
Throat infection - pharyngitis
Diagnosis and antibiotic combination*
Difficulty swallowing liquids
Difficulty swallowing solids
Pain on swallowing
Pain or burning when swallowing
Pain with swallowing
Cough due to angiotensin-converting enzyme inhibitor
Cough secondary to angiotensin converting enzyme inhibitor (ACE-I)
Problems with swallowing
Cryptogenic organizing pneumonia
DOE (dyspnea on exertion)
Swallowing pain or burning
Dyspnea on exertion
Other dyspnea and respiratory abnormality
Recurrent upper respiratory infection (URI)
Respiratory tract infection
Shortness of breath
Shortness of breath dyspnea
Shortness of breath on exertion
SOB (shortness of breath)
URI (upper respiratory infection)
DOE (dyspnea on exertion)
The development team next looked at which patient-specific elements of the history and physical exam (auto-generated when possible) the tool could use to automatically calculate the risk probabilities and provide recommendations suggested by validated CPRs. While several alternatives including traditional CDS templates were considered, it soon became clear that dynamic flowsheets would be used because this functionality would enable the required calculations of CPRs while maintaining the hub-and-spoke linkages critical to successfully integrating CDS tools into workflow .
Bundled order sets, documentation, and patient instructions
The study was conducted at a large urban academic medical center. All of the providers were members of the academic primary care practice that is located on the main hospital campus. The outpatient clinic has over 55,000 visits annually and serves a diverse population that is approximately 56% Hispanic, 35% African-American, 7% white and 2% other.
Provider eligibility, consent, and randomization
All primary care providers within the medical practice were eligible for the study. The practice includes 149 primary care faculty, residents, and nurse practitioners divided into four units on the same floor. The study design was a randomized control trial in which the providers within the academic medical center outpatient practice were the unit of randomization. Faculty providers were randomized via random number generator to intervention or control in a 1:1 ratio. Medical residents were randomized within blocks according to their outpatient ambulatory care month (a period with substantially increased outpatient clinical activity) assignments to ensure even distribution throughout the academic calendar. However, due to changes in the resident calendar in year two of the study, any additional medical resident providers entering the system were added in a 1:1 fashion. Only providers randomized to the intervention are triggered by the EHR to use the iCPR tools. After randomization, all providers were invited to standardized educational forums for consent and training (if randomized to the intervention).
All providers allocated to the intervention received approximately 45 minutes of training on how the iCPRs are integrated into the EHR and how to interpret the output of each iCPR. Each training session was led by at least one study investigator and one study staff member. The training consisted of a background on the strep throat and pneumonia CPR evidence, several walkthroughs of iCPR tools using the EHR training version, and a demonstration video simulating the tool in a live clinical encounter. Providers who were unable to attend group training sessions were trained individually.
Patient inclusion and exclusion criteria
There was no specific patient inclusion/exclusion criteria used in iCPR. The initial plan had been to use age, prior hospitalization history, and current/recent antibiotic use as criteria, but these were eliminated due to a variety of reasons, including inadequate/inaccurate documentation of prior medical history and current medication prescription. Thus, other than being an enrolled intervention provider, the only criteria for inclusion were the appropriate triggering diagnoses, chief complaint, or a diagnosis/order combination. The list of chief complaints and related diagnoses and orders that trigger each iCPR is listed in Table 1. Common triggers include a chief complaint or diagnosis of 'sore throat' for the strep throat iCPR and a diagnosis of 'bronchitis' for the pneumonia iCPR.
Patient characteristics, including age, gender, comorbidities, smoking history, recent hospitalizations and current or recent medications, are captured via EHR chart review.
Provider characteristics including age, gender, and years of practice are captured via self-report.
% of eligible encounters accepting iCPR and using bundled order set
Number of antibiotics prescribed
Number of chest x-ray ordered
% of eligible encounters accepting iCPR and using bundled order set
Number of antibiotics prescribed
Number of rapid strep tests and throat cultures ordered
The outcome measurement battery is designed to detect changes in clinical practice that are most likely to result from use of the iCPRs. The primary outcome is the difference in antibiotic prescribing frequency among patient encounters eligible for the iCPR tool among intervention compared to control providers. For example, for all patients presenting with symptoms that launch the pneumonia or strep throat tools, data will be collected from the EHR on the number of prescriptions for antibiotics written by providers randomized to the iCPR compared to usual-care arms, respectively. We will also examine the rate of chest x-ray orders and rapid strep throat test orders between intervention and control providers as a secondary outcome (see Table 2).
Data monitoring and quality control
All data collection is conducted via the EHR. Weekly reports are generated to track the frequency of the tool triggering, including the use of each component of the iCPR tools and the respective diagnostic triggers. Periodic chart reviews are conducted to monitor the appropriateness of tool triggering and to investigate any concerns raised by providers regarding usability or workflow disruptions. In addition, provider refresher training is conducted prior to residents coming onto each subsequent ambulatory care block in order to maintain a consistent ability to use the tool. The refresher consists of a videoclip simulation of a provider and patient interacting with tool.
The planned statistical analyses include comparing socio-demographic and other baseline characteristics. Patient comparisons will be conducted by stratifying the sample by randomization status and by condition (i.e., pharyngitis and pneumonia). We will use the t-test, Wilcoxon test, or the chi-square test, as appropriate, to evaluate the balance between groups. The relative frequency of triggering of each iCPR in intervention and control patients, overall, and by where the triggering occurs (chief complaint, ordering, or diagnosis) will then be compared. We will calculate the proportion of intervention encounters in which each component of the iCPR tool, including the overall tool, the risk calculator, and the bundled order set, are used. This calculation will be repeated stratifying by test condition and by provider characteristics (training level, et al.). To test the effect of iCPR, we will use a generalized estimating equation model with clinician as the cluster variable, antibiotic prescribing as the outcome variable, and intervention group as the only explanatory variable. Given the nature of the possible relationship between patients in a cluster, we will use an exchangeable correlation structure for parameter estimation.
Sample size was calculated as if individuals were independent, and then adjusted to account for the clustering of patients within physicians. Although patient outcomes are assumed to correlate somewhat within provider, the multicausal nature of clinical outcomes and the likely random nature of patient assignment to providers led us to estimate a small interclass correlation (intraclasss correlation coefficient for binomial response < 0.15). The calculation of sample size was performed with a significance level of 0.05 and 80% power. The adjusted sample size was calculated by multiplying the initial estimate of the number of patients by an inflation factor, which is a function of the interclass correlation and the number of the clusters. Final calculations estimated a need of 1,070 study subjects (535 in each disease condition) in total assuming a baseline rate of 30% antibiotic ordering in each condition and an estimated effect size of a 12% reduction in ordering in the intervention arm.
Several steps were taken to ensure a smooth and successful implementation of the iCPR CDS. A rapid response team composed of informatics and clinical expertise was available via pager for the first week after roll-out to respond to early bugs and other issues in real time. In addition, the team later embedded an option into iCPR for users to send messages to the build team to communicate issues. Furthermore, the lead clinician maintained a 'presence' in the practice so that any building frustration or problems with the tool could be handled rapidly before it built into more substantial resistance. Lastly, periodic focus groups were held to elicit users' feedback on the tool; these data were used to conduct ongoing refinements. The study was launched in December 2011 and is ongoing.
The iCPR trial was designed to assess whether a highly integrated CDS tool that supports clinicians in making EBM guided decisions is feasible, accepted, and effective. The team composition and design choices throughout the development process reflect the project's focus on enhancing provider acceptance and usability. The tool was designed by a multi-disciplinary development team that encouraged clinician users and designers to work together from inception. Iterative, in vivo usability was another key towards enhancing clinician acceptance because the think aloud and trained actor 'live' simulations each provided feedback that substantially improved the prototype. This approach differs from the more traditional usability testing under carefully controlled conditions that often minimizes the input of actual users in a realistic use setting . Standardized training demonstrated the new workflows to all intervention clinicians; another likely contributor to broad acceptance of the tool. Too often, new tools are rolled out into production with suboptimal training, creating resistance among providers . In summary, we believe that this 'grassroots' approach paired with usability and user training will improve previously disappointing update of similar CDS tools . The overall acceptance of the tool and its ability to alter antibiotic prescribing for suspected strep or pneumonia will be determined by the final outcomes of the trial. However, the approach used serves as a model for a more user-centered design of CDS; one that maximizes provider input and likely acceptance. These lessons should be generalized more broadly in CDS development of EBM and other point-of-care CDS tools.
Agency for Health Care Quality and Research (AHRQ) - 7R18HS018491-03
- Brook RH: The End of the Quality Improvement Movement. JAMA: The Journal of the American Medical Association. 2010, 304 (16): 1831-1832. 10.1001/jama.2010.1555.View ArticlePubMedGoogle Scholar
- Tonelli MR: The limits of evidence-based medicine. Respir Care. 2001, 46 (12): 1435-1440. discussion 1440-1431PubMedGoogle Scholar
- McGinn TG, Guyatt GH, Wyer PC, Naylor CD, Stiell IG, Richardson WS: Users' guides to the medical literature: XXII: how to use articles about clinical decision rules. Evidence-Based Medicine Working Group. JAMA. 2000, 284 (1): 79-84. 10.1001/jama.284.1.79.View ArticlePubMedGoogle Scholar
- Wang SJ, Kuperman GJ, Ohno-Machado L, Onderdonk A, Sandige H, Bates DW: Using electronic data to predict the probability of true bacteremia from positive blood cultures. Proc AMIA Symp. 2000, 893-897.Google Scholar
- Walsh BT, Bookheim WW, Johnson RC, Tompkins RK: Recognition of streptococcal pharyngitis in adults. Arch Intern Med. 1975, 135 (11): 1493-1497. 10.1001/archinte.135.11.1493.View ArticlePubMedGoogle Scholar
- McGinn TG, Deluca J, Ahlawat SK, Mobo BH, Wisnivesky JP: Validation and modification of streptococcal pharyngitis clinical prediction rules. Mayo Clin Proc. 2003, 78 (3): 289-293. 10.4065/78.3.289.View ArticlePubMedGoogle Scholar
- Heckerling PS, Tape TG, Wigton RS, Hissong KK, Leikin JB, Ornato JP, Cameron JL, Racht EM: Clinical prediction rule for pulmonary infiltrates. Ann Intern Med. 1990, 113 (9): 664-670.View ArticlePubMedGoogle Scholar
- Shortliffe E, Cimino J, eds: Biomedical Informatics: Computer Applications in Health Care and Biomedicine. 2006, Springer, ThirdGoogle Scholar
- Linder JA, Schnipper JL, Tsurikova R, Yu T, Volk LA, Melnikas AJ, Palchuk MB, Olsha-Yehiav M, Middleton B: Documentation-based clinical decision support to improve antibiotic prescribing for acute respiratory infections in primary care: a cluster randomised controlled trial. Inform Prim Care. 2009, 17 (4): 231-240.PubMedGoogle Scholar
- Mann D: Making Clinical Decision Support More Supportive. Medical Care. 2011, 39 (2): 115-116.View ArticleGoogle Scholar
- Schnipper JL, Linder JA, Palchuk MB, Yu DT, McColgan KE, Volk LA, Tsurikova R, Melnikas AJ, Einbinder JS, Middleton B: Effects of documentation-based decision support on chronic disease management. Am J Manag Care. 2010, 16 (12 Suppl HIT): SP72-81.PubMedGoogle Scholar
- Linder JA, Ma J, Bates DW, Middleton B, Stafford RS: Electronic health record use and the quality of ambulatory care in the United States. Arch Intern Med. 2007, 167 (13): 1400-1405. 10.1001/archinte.167.13.1400.View ArticlePubMedGoogle Scholar
- Welch WP, Bazarko D, Ritten K, Burgess Y, Harmon R, Sandy LG: Electronic health records in four community physician practices: impact on quality and cost of care. J Am Med Inform Assoc. 2007, 14 (3): 320-328. 10.1197/jamia.M2125.View ArticlePubMedPubMed CentralGoogle Scholar
- Sittig DF, Wright A, Osheroff JA, Middleton B, Teich JM, Ash JS, Campbell E, Bates DW: Grand challenges in clinical decision support. J Biomed Inform. 2008, 41 (2): 387-392. 10.1016/j.jbi.2007.09.003.View ArticlePubMedGoogle Scholar
- Persell SD, Kaiser D, Dolan NC, Andrews B, Levi S, Khandekar J, Gavagan T, Thompson JA, Friesema EM, Baker DW: Changes in performance after implementation of a multifaceted electronic-health-record-based quality improvement system. Med Care. 2011, 49 (2): 117-125. 10.1097/MLR.0b013e318202913d.View ArticlePubMedGoogle Scholar
- Shah NR, Seger AC, Seger DL, Fiskio JM, Kuperman GJ, Blumenfeld B, Recklet EG, Bates DW, Gandhi TK: Improving acceptance of computerized prescribing alerts in ambulatory care. J Am Med Inform Assoc. 2006, 13 (1): 5-11. 10.1197/jamia.M1868.View ArticlePubMedPubMed CentralGoogle Scholar
- Kushniruk AW, Borycki EM: Low-cost rapid usability engineering: designing and customizing usable healthcare information systems. Healthc Q. 2006, 9 (4): 98-100. 102PubMedGoogle Scholar
- Ash JS, Chin HL, Sittig DF, Dykstra RH: Ambulatory computerized physician order entry implementation. AMIA Annu Symp Proc. 2005, 11-15.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.