Referral data metrics: What is lacking and how to improve data practices going forward

Our previous blog, “Referral data metrics: what they are, what exists, and why they are important,” outlined the five data metrics that inform navigation systems: quality control metrics, demand metrics, supply metrics, referral metrics, and system impact metrics. If you haven’t read the previous entry, check it out first because this article extends this discussion of referral metrics. 

In this article, we describe essential improvements to metrics that are not currently implemented in systems of care. These improvements are necessary because they help network leaders and funders determine what is working and scale those practices. In addition, they help protect clients from practices like skimming, where partners select the clients with the least challenging needs to improve their agency’s scores.  “Mapping the Navigation Systems of Pennsylvania: Opportunities for The Future” highlights four areas where current referral metrics fall short.

1. Referral metrics are not tracking what aspects of referrals result in reduced costs and greater client wellness.

Emergent research has indicated that navigation systems reduce the cost of care. But, it remains unclear which specific practices navigation systems are reducing this cost. This knowledge gap requires specific data metrics that are not currently available. 

Some potential questions that navigation systems might ask themselves:

  • Is the use of screeners correlated to reduced longitudinal costs?
  • Does the coordination of multiple types of care decrease collaboration costs?
  • Does the personal relationship formed with human navigators reduce cost by lessening interactional costs?
  • Do knowledge of benefits, services, and programs decrease the cost of locating and navigating care? 

These questions and others similar critically attempt to get to the bottom of the cause of this reduced cost. This knowledge could inform future decisions that would further reduce costs and increase the adoption and usage of referral platforms generally.

2. Referral Metrics are not tracking what aspects of referrals result in greater client wellness.

Navigation systems are currently tracking metrics such as how fast a client receives services and the call quality between the navigation system and the client. While these metrics are helpful in their own right, they do little to tell us whether or not the client is well due to the interaction or the call. Other data metrics are needed to answer the essential question of whether a client’s wellness improved due to the use of navigation systems. 

Some potential questions that navigation systems might ask themselves:

  • Does the use of screeners promote more holistic wellness?
  • Are clients who use navigation systems collectively better off than those who do not?
  • Are the services provided by navigation systems conducive to increased client wellness?
  • Are there any practices that are decreasing client wellness?

The end goal of navigation systems is to ensure that clients receive the help they require and, by extension, the wellness they seek. Tracking the sources of increased wellness helps to inform the wellness of all future clients of navigation systems.

3. Referral metrics are rarely disaggregated by race/ethnicity.

State-of-the-art navigation systems track accuracy, efficiency, and service-episode outcome. This tracking allows network leaders to examine outcomes by both service type and provider. But, this outcome data is rarely disaggregated by race, ethnicity, discharge status, zip code, or other client attributes. 

Some potential questions that navigation systems might ask themselves:

  • How do outcome accuracy, efficiency, and status differ based on zip code, race, and identity?
  • Are English-speaking clients receiving better care than non-English-speaking clients?
  • Are specific backgrounds having cases closed out less often than others (e.g., other than honorably discharged veterans)?

More specific outcome data might indicate structural and institutional racism within navigation systems. It may also indicate other forms of unbalanced treatment. Identifying such bias would be the first step to rectifying the problem and ensuring equal access to care.

4. Referral metrics lack predictive analytics to identify the best trajectories of care.

The social determinants of health suggest that clients seeking help typically have groups of needs rather than particular needs. Navigation systems regularly track client referral data longitudinally for service episodes of a singular need. But, referral systems do not currently track the best trajectories or sequences of care for multiple needs. This episode-centric view is problematic because a client might need food, housing, and employment but just receive food assistance. Human navigators might be puzzled when this same successful client returns later to ask for food once again. And, not only are clients’ needs typically multidimensional, but navigation networks might best serve these needs in a specific sequence. For example, research has emerged indicating that addressing housing needs before other human service needs results in better outcomes (e.g., Housing First). There may be other similar sequence patterns that optimize care.

Predictive analytics is a critical intervention that relies on knowing the best trajectories of care. Predictive analytics provide information that helps indicate and address a problem before it occurs. It does this by flagging specific indicators that suggest future problems. One present example of this is Hello Baby in Allegheny County, PA, where predictive analytics use predictive analytics to identify families that could use more significant support, based on their history of social service utilization and interaction with criminal justice systems. The goal of this program is to prevent the need to involve Child Protective Services through early intervention. Examples like Hello Baby show the value of getting involved earlier than a client asks for help.

Some potential questions that navigation systems might ask themselves:

  • What are common clusters of client needs?
  • What sequence optimally addresses these needs?
  • What are common indicators of a need for additional help?
  • How can we intervene to prevent acute care needs through the use of predictive analytics?

Understanding and predicting the best trajectories of care could be the next significant advancement of referral platforms. This understanding could unlock more tailored and specific client-focused care that may have resounding effects on the wellness of clients longitudinally.

Data metrics are undeniably valuable, and their proper utilization in the service network landscape is critical to understanding how to provide care and how to provide care optimally. The new practices discussed in this blog serve as a jumping-off point to a better, more efficient, and more effective referral landscape.