Identifying and prioritizing patients for targeted interventions is a critically important component of an effective care management program. Direct patient intervention is a relatively costly service and program managers must identify the best candidates for the program in order to maximize clinical outcomes and return on investment for program sponsors.
Analyzing the patient population to better understand overall risks and comorbidities is necessary for projecting the potential impact on cost and utilization that a particular program might achieve. A common approach for assessing and stratifying patients for a disease management program is to use diagnoses and drug codes on claims data to identify patients being treated for targeted conditions and then apply risk adjustment models to determine the relative risk for each patient. Patients with higher risk scores are expected to incur higher costs, so these patients may be targeted first for more intensive interventions. While risk adjusters and relative risk scores can be very useful for care management programs this approach does not take into consideration the complex clinical mix of patients with comorbidities where one condition may be an obstacle to care management interventions targeting another condition. Another approach is to use a more clinically nuanced strategy and stratify patients based on clinical algorithms that do account for the interaction of chronic conditions, such as Milliman’s Chronic Condition Hierarchical Groups (CCHGs).
CCHGs is a patient centric analytic tool consisting of 43 non-overlapping categories. Twenty-four categories are for patients with chronic diseases and 19 categories for health patients (healthy is defined as the absence of a identified chronic condition). The classification rationale is to rank conditions based on how much they influence treatment plans.
The following example illustrates how CCHGs can be used to evaluate the potential of a disease management program designed to reduce admissions for Congestive Heart Failure (CHF). In a demonstration sample of data (refer to Table 1 below) there were 1,571 admissions for CHF (defined by Diagnostic Related Groups). If a program anticipates it can reduce the number of future CHF admissions by 10% (157 admissions), the program managers need to consider the fact that 18% of the CHF admissions in this population were for patients who have renal failure, about 11% of the admissions were for patients with active cancer and about 3.5% were for patients with severe psychiatric conditions. These non-CHF conditions are confounding conditions, meaning they interfere with or even prevent programs targeting CHF being effective. By using the CCHGs we can automatically remove these and other patients with confounding conditions and isolate those CHF patients where we expect to see an impact. In this example, if we subtract the number of admissions for patients with psychiatric conditions, renal failure and active cancer, the total number of target CHF admissions is 1,063. Ten percent of 1,063 equates to 106 admissions.
