Golden opportunities for clinical decision support in an era of team-based healthcare

Abstract

Computerized clinical decision support (CDS) will be essential to ensuring the safety and efficiency of new care delivery models, such as the patient-centered medical home. CDS will help empower non-physician team members, coordinate overall team efforts, and facilitate physician oversight. In this article, we discuss common clinical scenarios that could benefit from CDS optimized for team-based healthcare, including (1) low-acuity episodic illness, (2) diagnostic workup of new onset symptoms, (3) chronic care, (4) preventive care, and (5) care coordination. CDS that maximally supports teams may be one of biomedical informatics’ best opportunities to decrease health care costs, improve quality, and increase clinical capacity.

Introduction

Intractable challenges confront healthcare in the United States,¹ including unsustainable rising costs,^2-4 primary care physician shortages,⁵ physician burnout,⁶ and increasing demand for healthcare services.⁷ Primary care has been described as in crisis.⁸ Healthcare is too often inadequately standardized, of poor quality, and unsafe.^9,10

Some have suggested that health information technology (IT) holds the promise of driving improvements in healthcare.^2,11-13 Others have suggested that solutions to healthcare's challenges reside in the form of multi-disciplinary coordinated primary care, delivered through team-based approaches such as those found in patient-centered medical homes (PCMHs).^14,15 The most effective solutions to healthcare's critical challenges seem likely to arise from the synergistic combination of CDS and team-based approaches.¹⁶ Health IT’s (and CDS’) primary contribution may be in making new care delivery models possible.¹⁷

Clinical decision support for multi-disciplinary teams

Excellent opportunities exist for transforming primary care practice by fully leveraging the expertise of clinical teams. Both physicians and non-physician clinicians frequently practice below their level of licensure, with resultant underutilization of nurses and medical assistants.¹⁸ Pelak and colleagues estimate that over half of overall care performed by primary care physicians could be safely performed by nurses and other non-physicians.¹⁹ Similarly, Altschuler and colleagues estimate that 77% of time spent on preventive care and 47% of time spent on chronic care could be delegated to non-physicians, such as registered nurses, pharmacists, health educators, and medical assistants.²⁰

In the sections below, we describe common clinical scenarios and forms of computerized clinical decision support that could coordinate the parallel efforts of multi-disciplinary primary care clinical teams. We have categorized these primary care tasks into the following clinical domains: low acuity episodic illness, diagnostic work-up of new onset symptoms, chronic care, preventive care, care coordination and other tasks.

Primary care tasks performed by teams utilizing CDS

Clinical domains that could benefit from CDS optimized to support teams include:

Low-acuity episodic illness. As demonstrated by care delivered at an increasing number of retail clinics, nurse practitioners or physician assistants employing guideline-based protocols can effectively manage common low-acuity episodic illness, such as sinusitis and pharyngitis.^21,22 Compared with primary care clinics, there is evidence that retail clinics can deliver this healthcare at lower cost with equivalent quality.^21,22 In one study, ten conditions commonly treated at retail clinics accounted for an estimated 18 percent of all primary care physician visits.²³

Primary care clinics could readily adopt similar guideline-based methods to manage low-acuity episodic illness using computer-based standing orders protocols administered by nurses. An advantage that primary care clinics would have over retail clinics is improved capacity to evaluate persistent symptoms, which in some cases represent serious underlying disease. Computer logic could route these persistent cases to the physician if a scripted telephone follow-up administered by a medical assistant indicated that the symptoms did not resolve or respond as expected.

Diagnostic workup of new onset symptoms. For patients who present with symptoms suggestive of more serious underlying disease (e.g., hemoptysis), computer-based scripted methods might provide valuable first-pass diagnostic workup immediately prior to physician evaluation. Through computer-based standing order protocols, preliminary diagnostic tests could be performed prior to the physician's interview of the patient - e.g., a chest x-ray for hemoptysis in an older smoker who has not previously had chest imaging.

Using structured questionnaire data acquired by non-physicians (including from the patient directly) that relate to the patients' symptoms, computer logic could generate a draft differential diagnosis and draft treatment plan for physician review available as the physician goes into the exam room. Such a draft differential diagnosis could list relevant physical exam findings or diagnostic testing that could further narrow the range of diagnostic possibilities.

The availability of first-pass diagnostic approaches acquired through non-physician personnel has the potential to improve upon the estimated 10 to 15% diagnostic error rate.²⁴ Such efforts might also finally provide the means to insert differential diagnosis generators, such as DXplain²⁵, into routine clinical workflow. In turn, draft treatment plans generated by questionnaire data could incorporate recommendations that reflect the best and latest medical evidence, potentially reducing the long delays between definitive research findings and widespread practice.²⁶ In all cases, it would be the physician or other treating provider who would determine based on their exam if the draft differential diagnosis or treatment plan were appropriate for that particular patient.

Chronic care. Guided by electronic questionnaires, non-physicians could capture a much larger portion of history-taking and documentation related to chronic care than currently performed. For example, non-physicians could collect longitudinal patient data necessary for physician decision-making related to asthma therapy, such as the frequency of inhaler use, coughing, or episodes of shortness of breath. They could also capture structured data needed for chronic care quality measures.

To the significant extent that questions are predictable on the basis of a patient's presenting symptoms, the goal would be to use computer logic to collect the answers to those questions directly from the patient (e.g., using tablet technologies) or through a medical assistant.²⁷ For example, if a patient reports chest discomfort, questions regarding its location, severity, radiation, and association with shortness of breath could follow. Such capture of patient history could relieve the physician of significant portions of visit documentation, allowing the physician to instead focus on verifying the "pertinent positives" from the captured history, supplementing with any needed clarifying questions, and establishing a therapeutic plan with the patient.

For patients with problems identified by a physician, computer logic might also help guide the selection or titration of various medications related to chronic care conditions. An early example of CDS for medication selection in diabetes has been developed by Tarumi et al.²⁸ Such computer logic has been demonstrated as beneficial for anticoagulants and insulin compared with routine care.²⁹ There is also demonstrated utility for computer-based hypertension management.³⁰ Logic could generate medication recommendations to be approved by the physician, or eventually with the proper safeguards, computer-based standing order protocols could directly guide medication titration.

With safeguards in place, a physician might identify a patient in the future with a new diagnosis of hypertension or in need of improved glucose control. That physician could then choose to turn over the choice of medication and/or the subsequent medication titration to non-physicians using computer-based standing order protocols. For purposes of choosing the “right” anti-hypertensive agent, the computer logic itself could search for evidence that the patient has a history of systolic heart failure, diabetic nephropathy, or recent myocardial infarction.

Preventive care. Through computer-based standing order protocols, non-physicians could directly coordinate much more preventive care than commonly occurs. In the case of vaccinations, we have previously demonstrated that automated standing order protocols are more effective than physician-directed computer reminders.³¹

Using these approaches, the default for a particular clinic might be that age-specific preventive care is offered to all patients, unless and until such time that the patient in coordination with their physician chooses to stop (e.g., choosing not to proceed with screening colonoscopy due to terminal illness). In cases where the patient refuses recommended preventive care, the reason could be captured electronically and routed automatically to the physician, on the possibility that additional patient education would be useful.

Rather than spending as much time directly ordering preventive care for individual patients, physicians could instead focus on practice-level oversight. There is much potential for saving physician time through such methods given estimates that 7.4 clinician hours per working day are needed to fully follow national recommendations for an average-size panel of 2,500 patients.³²

Care coordination and other tasks. Compared to the case where a single clinician is responsible for all orders and documentation, team-based primary care does complicate issues of care coordination. Consequently, the potential value of CDS software is arguably much greater in a team-based environment than in the simpler single physician case. The goal would be that regardless of how work is dynamically allocated or delegated within a practice, or regardless of the sequence by which tasks are completed, there would be software verification that all aspects of healthcare are addressed by the end of the visit.³³

Apart from the clinical scenarios described above, many other activities occur frequently in primary care clinics³⁴ that could benefit from team-based CDS software:

• Establishing the patient’s agenda and priorities for the imminent physician visit

• Medication reconciliation

• Reviewing potential side effects of newly prescribed medications with the patient

• Follow-up of hospitalizations and emergency department visits employing standardized questionnaires with forwarding of summaries to the physician.

• Renewals of non-narcotic medications using electronic standing order protocols

• Triage of returning test results, incoming emails and phone calls, insurance clarifications.

• Tracking of referrals from the original order to returning recommendations from the specialist

• Follow-up of missed clinic appointments or diagnostic testing

Team‐based approaches hold the key to allowing physicians to spend more of their time overseeing the delivery of healthcare within their practice, rather than personally delivering the vast majority of healthcare. To fully take advantage of the opportunity for physician supervision of practice-level information, much would need to be learned regarding how to optimally summarize the results of team efforts. Valuable reports might include both summary and patient-level data on (1) preventive care administered or refused, (2) blood pressures before and after changes in treatment, (3) medication titrations, and (4) the frequency and nature of various concerning patient symptoms.

Discussion

In team-based care models, CDS software has the potential to be invaluable, transforming potentially fragmented team efforts into highly effective integrated healthcare delivery. Despite risks of fragmented care, the clinical advantages of multi-disciplinary medical care will likely drive wide adoption of team-based practice. CDS that maximally supports teams may be one of biomedical informatics’ best opportunities to decrease healthcare costs, improve quality, and increase clinical capacity.

In this paper, we have outlined clinical tasks that a diverse group of clinicians might effectively address using CDS. Collectively, low-acuity episodic illness, diagnostic workup, chronic care, preventive care, and care coordination account for a large portion of medical practice. Many issues would need to be worked out. For example, full adoption of team-based care approaches might require changes to reimbursement methods, rewarding practices for overall care delivered rather than who delivered what care.

Practices also differ in their precise assortment of team members, so CDS software would need to be highly configurable in terms of which team members are assigned particular tasks. Depending on the practice, available non-physician team members might include nurses, physician assistants, pharmacists, medical technicians, rehabilitation specialists, social workers, check-in personnel, or patient navigators. In a number of cases, more research would be required prior to wide deployment, such as ensuring that adjustment of anti-hypertensive or diabetic medications employing computer-based standing orders is safe and effective.

Our suggested approaches take advantage of the fact that a significant portion of healthcare is algorithmic in nature, even for the preponderance of cases where “the art of medicine” must subsequently be applied. In a handful of cases, some interventions and clinical scenarios are sufficiently straightforward that they can be safely administered by non-physician clinicians with only high-level provider oversight (e.g., flu shots³¹). For most clinical scenarios, however, the goal would be to simply provide the clinician succinct diagnostic and therapeutic considerations that they can review at the point of care. Ready availability of such considerations at the point of care would best be complemented by trivial methods to directly order those same diagnostic tests or therapies.

The shortcomings of EHR support for team-based care have been previously recognized.^2,16,35-38 The American Medical Association has called for EHR software that (1) facilitates clinical staff to perform work to the extent their licensure and privileges permit, and (2) allows physicians to dynamically allocate and delegate work to appropriate members of the care team.³⁵ Bates et al. recommends the development of EHR functionality to enable real-time communication and coordination among team members.¹⁶ O'Malley et al. suggests that EHRs could facilitate primary care teamwork by enhancing communication within the practice team, supporting task delegation, and integrating standing orders and protocols. In this paper, we identify specific primary care tasks that could be performed by non-physicians supported by software. We believe these specific examples strengthen the already-compelling case for improving team support in CDS software.

Team-based approaches have been associated with increased physician satisfaction and decreased burnout.^39,40 Too many PCPs practice within a "frantic bubble," experiencing workdays as a non-stop stream of patients.⁴¹ Almost all aspects of a primary care physician's busy day could benefit from team‐based assistance. CDS software can make that possible.