BY Ayal Bitton, Advisory Managing Director, and Alex Obenauf, Advisory Managing Director, Customer and Operations Health & Government, KPMG
For many provider organizations, one of the most challenging aspects of centralizing patient scheduling is systematizing accurate appointment scheduling. In decentralized (practice) settings, providers’ diverse preferences for how, when and how long a patient ought to be scheduled are often accommodated by localized knowledge and proximity to providers. The scheduling of the same service within the same specialty can differ depending on the provider. For organizations seeking the efficiency and service predictability of centralization, importing the individual scheduling preferences of hundreds or thousands of providers into a centralized setting is daunting.
Further complicating the challenge is the notion that patients’ unique clinical circumstances should drive different 
scheduling outcomes. For instance, a healthy 25-year-old patient seeking a physical from her primary care provider may require less time for the same service than an 81-year-old patient with hypertension, diabetes, anxiety, and gout. Scheduling processes that are already subject to significant provider customization take on infinitely more permutations depending on the clinical and psychosocial needs of patients.
Factoring patients’ clinical complexity into scheduling algorithms at scale has been elusive with legacy technologies
Historically, many organizations have sought to accommodate diverse provider scheduling preferences with catalog-like knowledge management tools, ranging from less sophisticated binders of scheduling guidance to more digitally navigable versions of the same. In the last several years, guided scheduling logic and decision trees have become a central feature of more sophisticated approaches to accurate and streamlined scheduling, based on the premise that like services among like providers ought to be similarly scheduled. Epic electronic medical record’s decision tree capabilities in its Cadence module have gone further in reducing scheduler reliance on rote question-and-answer patient interactions by leveraging rules-based logic inter-spliced with questions that expand the potential permutations of relevant factors while limiting the number of (often redundant) scheduling questions to patients. Rules evaluating patient history and attributes run in the background to capture accurate information — invisible to schedulers and patients — surfacing only relevant patient questions based on patient data, eliminating, for example, the need to ask every patient whether they are new or established. Despite all these advancements, however, there have been few opportunities to incorporate clinical complexity into guided scheduling logic without risking the prospect of clinical decision-making or interpretation by nonclinical scheduling staff.
Leveraging risk stratification tools and guides scheduling to incorporate clinical complexity in patient scheduling
KPMG recently overcame this challenge with a client by incorporating into Epic decision tree logic the use of Johns Hopkins Adjusted Clinical Groups (ACG) and hierarchical condition category (HCC) coding, central features of the client’s quality and population health management programs. These conventions stratify patient risk based on various demographic, social, and diagnosis-based risk factors.
Having facilitated a collaborative process to optimize their guided scheduling logic, KPMG and client stakeholders conducted independent scenario-based scheduling evaluations with providers and leaders to determine risk-scoring alignment with proposed scheduling logic. KPMG professionals and client stakeholders determined that patients with certain risk profiles could be allotted additional time for select primary care visits without diminishing provider efficiency or asking nonclinical schedulers to evaluate patients’ broader health status. The guided scheduling algorithms programmed to contemplate risk scoring facilitated a standardized approach to patient scheduling while predictably allowing providers additional time to treat their most clinically complex patients, all without customized provider scheduling guidance.
This effort was a win-win for centralized scheduling operations and provider satisfaction. After incorporating additional time for visits based on ACG and HCC complexity, the health system still realized a 5-minute decrease in average visit duration by standardizing appointment lengths and provider template structures, resulting in about a 10% increase in visit capacity. A defined clinical complexity score is more efficient and optimal than individual scheduling practices.
Conclusion
Accurate patient scheduling is dizzyingly complex for reasons that span the diversity of providers and their practices, and the clinical complexity of individual patient needs. Solving these challenges systematically and at scale is more important now than ever given increasing market demands for convenient access, intensifying top-line and margin pressure, and the underlying imperative to deliver care more efficiently. The tools and technology to solve these challenges are more capable than ever. The challenges in implementing them, however, are not uniquely technical.