Given the healthcare industry’s pursuit of transitioning to value-based care, the need to better manage patients’ underlying conditions and minimize potentially avoidable costs is growing faster than ever. Care management is now important for many more individuals than the usual high-risk, high-touch patients—but more effective programs need to be developed. While some current care management programs have produced success stories, most are not showing a meaningful impact on the patients whose health they are attempting to improve, and thus they are leaving significant value (in terms of both outcomes and costs) on the table. One study, for example, determined that typical engagement rates for a health insurer’s disease management programs averaged only 13 percent.
Our experience is similar. We have found, for example, that for patients at risk for an unplanned readmission after an initial inpatient stay, some care management programs are leaving as much as 95 percent of the potential impact unaddressed. Too often, these programs do not identify the right people to target, fail to reach out to those people successfully, and cannot engage patients in the ways needed to prompt behavior changes (Exhibit 1).
Payers and providers can achieve the full potential of care management—including a positive ROI (often, above 2:1)—if they substantially improve their ability to identify, reach out to, and engage patients. In our experience, technological advancements, including the evolution of big data, decreased cloud computing costs, and improved machine learning capabilities, have created an opportunity for care management to succeed at levels not previously possible.
Within a care management program, two types of advanced analytics can be used. Descriptive analytic approaches look at past patient data to help identify sources of value—for example, avoiding complications, choosing the right setting of care, or shifting care to high-performing providers. Predictive analytic models, on the other hand, forecast future trends such as what paths patients are likely to take; machine learning makes it possible to continuously create better predictions of what patient path is most likely. Thus, descriptive analytics makes it possible to determine which patients are at risk of unplanned readmissions; predictive analytics can then be used to identify the patients with the highest probability of potentially avoidable medical costs—and then alert care managers to intervene at the right time. Just because a patient incurs high costs today does not mean that he or she will continue to incur high costs in the future, or that those costs will be avoidable. Predictive models that focus on future avoidable costs, rather than simply observing current high costs, can help payers and providers determine if they need to target their care management resources on a different set of patients.
More specifically, detailed patient journey analytics can be used in care management programs to spot patterns associated with poorer outcomes and increased costs (Exhibit 2). Among patients with chronic obstructive pulmonary disease (COPD), for example, our model identified medication nonadherence as a strong predictor of a future avoidable emergency department (ED) visits. We then used machine learning algorithms to segment and prioritize patients based on the likelihood that they would be willing to change their behavior to improve their health and thus lower the risk of a high-cost event. We found that one pattern of behavior—COPD patients who consistently filled their medications initially but then stopped refilling them for about three months—was linked to an especially high risk of having an exacerbation that involved ED care. We were then able to design an intervention for these patients to help them avoid an exacerbation.
Once patients are identified and specific interventions are designed, the care management program should use consumer-based approaches to identify the delivery model that best meets the patients’ preferences and maximizes the likelihood that the program engages them. The delivery model should include:
- The medium for communication (e.g., phone calls, emails, app notifications, videos, text messages)
- The message
- The pattern of outreach (e.g., time of day, frequency of follow-up requests)
- A strategy for matching patients with the right care manager
Analytics can improve development of the delivery model by measuring and segmenting patients based on previous responses to various methods of delivery—in essence, determining which methods are most likely to prompt behavior change. For example, instead of sending all program participants text messages regardless of their preferences and risk levels (as is often done today), analytics can determine which high-risk individuals would prefer and benefit from in-person outreach via video chat. This example ultimately creates a higher ROI despite the higher-touch involvement.
Digital approaches to engagement
Once analytics have helped identify the types of interventions most likely to be effective, digital capabilities can help maximize the effectiveness of engagement by fine-tuning the message that goes with the interventions, as well as the timing, frequency, and channel for outreach. Although digital approaches do not necessarily change the underlying care interventions, they significantly broaden the mechanisms for delivering care management interventions beyond the traditional means (which have generally been limited to nurse calls or nurse visits). By coupling analytics and digital approaches, payers and providers can not only identify who to target, when and how, but also maximize the effectiveness and efficiency of delivering the interventions.
Patients are increasingly preferring digital solutions: In our 2018 Consumer Health Insights survey, 75 percent of the respondents said they prefer digital solutions for monitoring their health, and 71 percent preferred digital solutions for checking their health data.
Personalization is also becoming critical; an overwhelming majority of US consumers now expect companies to understand their individual needs.
Crucially, digital outreach methods allow for rapid and cost-efficient experimentation. For instance, in an A/B testing approach, two groups of people are engaged in slightly different ways (for example, text messages versus app notifications, emails with different subject lines, and text messages with different wording). The version that is more effective in getting a targeted set of recipients to engage is then selected for broader use. A/B testing has long been used in other industries, spearheaded by companies such as Amazon
; we believe this approach holds promise in a range of healthcare settings, including care management.
Digital capabilities can also help drive down costs in care management programs. In our experience, companies can achieve a 10 to 20 percent decrease in administrative costs to deliver the same level of outreach. For example, digital approaches can be used to provide prescriptive guidance to nurses on the scope of care management programs and to automate many manual tasks (e.g., dialing patients, speeding up assessment and note-taking). As a result, the clinical care management staff can maximize the time spent on patient-facing activities.
Several healthcare organizations that have started to embrace the full capabilities of analytics and digital are realizing significant savings. For example, one health insurer that used these approaches to transform its care management programs achieved a 2:1+ ROI
; another organization realized a 25- to 30-percent reduction in ED visits.
Admittedly, some of the needed changes require significant time and investment to implement. However, there are also a series of changes that care management organizations can begin implementing today because they require little to no additional investment.