When it comes to risk stratification — the process of identifying the right level of care and services required by a distinct group of patients — medical group practices need an effective model to get the most out of value-based reimbursement. A one-size-fits-all approach, according to Jeanette Ball, BSN RN, PCMH CCE, client solution executive, CTG, may not meet the needs of higher-risk patients, including those with chronic conditions such as diabetes, hypertension or kidney disease.
“Make sure that you have broken this stratification down into groups that help you make sense of the whole by looking at the smaller pieces,” explains Ball regarding segmenting patients into distinct groups of similar complexity and care needs. “And then design interventions to improve those outcomes.”
The EHR’s importance in building a risk model
To determine risk stratification and develop an effective risk model, it’s imperative to assess large amounts of data through your EHR. Claims data and data from risk-bearing organizations such as accountable care organizations (ACOs) can also be used, though that data isn’t as valuable as EHR data. According to Alan Mitchell, executive director, HealthEfficient, “Your EHR contains the most comprehensive, timely and accurate data for your patients,” while data from other sources can be used to fill in gaps. By using all relevant data, practices can improve clinical outcomes and close care gaps, especially for high-risk patients.
When reviewing EHR data, there’s much to consider including lab data, data from encounters, data from other healthcare providers and perhaps most important, data tied to social determinants of health (SDoH), which encompasses socioeconomic factors, physical environment, health behaviors and healthcare, for example. “This data is critical to understanding the total risk to your patients,” asserts Mitchell about the importance of pinpointing sources of data that can help practices succeed in a value-based arrangement.
Risk stratification tools
When considering risk stratification tools, Mitchell says practices need to assess their needs regarding commercially available software versus customized tools, depending on how the practice plans to generate risk scoring. “It’s possible you have risk scoring built into your [EHR] already,” notes Mitchell, adding that “an advantage of this is that if you’re comfortable with the risk stratification algorithm in your EHR, you will have much of what you need in a single platform.”
That said, Mitchell warns that practices should only attempt to build their own solution if there are significant gaps in what products already supply. He adds that in-house or contracted developers are necessary to design and maintain a customized solution. There are pros and cons to each, including those in Table 1.
Commercial risk stratification models
As noted by Mitchell, there are many evidence-based and well-established commercial risk stratification models available to practices. Mitchell says that practices should first review each of the leading models and “truly understand them, how they apply to your patient population, and whether they are feasible for your practice.” Although most population health analytics software packages typically integrate one or more of the models for an additional fee, the usability may make it worthwhile for practices. Some of the more prominent models are described in Table 2.
Mitchell says that an open source risk stratification model such as the CMS-HCC provides practices access to the elements that comprise the model’s algorithm. However, this is not the case with proprietary models such as the Johns Hopkins ACG, so practices should take this into consideration when deciding on a model.
If practices are just getting started with risk stratification, Mitchell points to the National Association of Community Health Centers’ (NACHC) model as one that is easy to implement and use because of its condition count, which is based on ICD-10-CM codes. On the other hand, if practices have experience with risk stratification models, Mitchell says that the American Academy of Family Physicians’ (AAFP) model is more comprehensive, with dozens of factors arranged into five primary categories, including potential physical limitations and utilization of healthcare resources.
“For this model to work, you will need to capture the data and compute it,” maintains Mitchell, adding that a “benefit of this model is that it goes well beyond clinical factors and considers the social determinants of health.” In addition, it also uses less-tangible qualitative factors that often aren’t easy to determine — for example, does the patient have a difficult time following a treatment plan and is the patient likely to be hospitalized in the next 30 days — yet important to include in a model.
Custom risk stratification models
Although a custom model may seem appealing because it can be structured to fit practice needs and patient population, Mitchell warns that the drawbacks often outweigh the benefits. “Such efforts are fraught with challenges, most especially the lack of a large evidence base to support whatever models you come up with,” says Mitchell of deciding to design your own model. Moreover, according to Mitchell, a custom model may provide unpredictable results and not give practices the ability to compare risk among different patient populations, which will prevent level setting.
Choosing the right risk stratification model
When choosing the right model for your practice, Mitchell suggests considering a number of key factors and conducting a trial run to determine fit, whether it’s a simple model that only takes into account chronic disease diagnoses and related factors or one that figures in SDoH as well. “Only your clinical leadership through the evaluation of each of these models and by defining and understanding your own practices’ and health systems’ goals can select the best model in consideration of the totality of circumstances,” says Mitchell of the importance of due diligence.
Key factors practices should consider when deciding on a risk stratification model:
- Access to needed data
- Ease of implementation in terms of practice and health information technology
- Relevance to patient population
- Visibility of its algorithm (if proprietary, it likely won’t be possible).
Once practices have assessed these factors, they should:
- Consult key providers and leaders
- Run a pilot with existing data to obtain insight on the value to the practice
- Account for other important measures such as SDoH, behavioral health, chronic conditions, and additional high-cost factors
- If a primary care practice, consider how the requirements of the Patient-Centered Medical Home (PCMH) standards apply to risk stratification, all with the goal of assessing value to the practice.
“If you can identify which patients are at the highest risk of adverse, high-cost outcomes, you can better align your resources to address those needs, therefore delivering value back into the system,” says Mitchell about how risk stratification models can benefit practices.