Social service systems succeed by pairing the right people with the right services. Yet too often individuals who could benefit from a social service receive a referral but never connect with the service.
In these cases, program staff must identify who is dropping off the referral pathway and where and when it happens. Agencies are often eager to identify problematic spots and develop plans to repair the pathway but may not know where to begin. Staff can use data to uncover bottlenecks, backlogs, or breakdowns in clients’ connection to and sustained engagement with services.
An initial data analysis can help agencies focus their efforts, especially before undertaking more resource-intensive diagnostic activities like case reviews, client shadows, and stakeholder interviews.
Three questions to start an effective data analysis:
What are the key referral steps and where are people dropping off? We often reduce a referral to a single step in the service delivery process — when, in reality, it is composed of many steps. Breaking down a referral process into individual steps allows us to more specifically identify where breakdowns may be occurring.
How long does each step take? Delays in the referral process can lead to breakdowns as clients become frustrated and disengage. Delays can also reduce the effectiveness of services if they mean the supports are no longer timely or responsive to client needs.
Who are we actually reaching? Not everyone will experience the referral process in the same way. We want to look for potential disparities in access to or progress through services by race, ethnicity, gender, geography, or other priority population characteristics.
What follows are some examples of how the Harvard Kennedy School Government Performance Lab has helped jurisdictions begin to answer these three questions. Keep in mind that some of the examples may answer more than one of the questions.
Question 1: What are the key referral steps and where are people dropping off?
Start by considering what steps someone would need to take to get connected to services. Common steps include identifying a need for a service, initiating a referral, connecting with a service, enrolling in the service, reaching key milestones, and finally, ending a service engagement. Jurisdictions may know more about who enrolls and who exits a particular service, but the in-between steps can feel like a black box. Identifying the steps and then tracking progress through each key step can shed light on significant challenges.
Make sure you identify all steps — not just yours and not just the ones you can easily see. Think about the actions and experiences of everyone involved: the person referring, the client, the service provider, etc.
New Hampshire Division of Children and Youth Services
Challenge: In New Hampshire, officials at the Division of Children and Youth Services and staff of Community-Based Voluntary Services were grappling with low program enrollment.
Data-driven Insight: Officials reviewed enrollment data and learned that 90% of families who enrolled did so after an in-person meeting with staff members. Yet, only 45% of families were seen in- person. Officials also knew that employees with lived experience were more successful at making initial contact with families.
Initial Results: After emphasizing thoughtful and intentional hiring and other intake strategies, the number of in-person meetings with families increased, which led to an increase in enrollment from 45% to 70%.
Washington Department of Children, Youth & Families
Challenge: In Washington state, officials at the Department of Children, Youth & Families (DCYF) knew pregnant individuals with substance use disorders were being referred to the child protection hotline yet screening out because there was no child present. These parents-to-be were not offered services, and many were later re-referred to the hotline, and the baby would often be removed for safety concerns.
Data-driven Insight: To reduce the number of infant removals, DCYF adjusted their referral process. Despite successfully referring 93% of screened-out substance- using pregnant individuals to community providers, only 33% of those individuals enrolled in a program or received another referral. When officials explored the low enrollment data, they learned community providers could only reach 37% of those being referred.
Initial Results: Officials saw that two providers had higher enrollment rates, so they explored ways to spread the seemingly successful techniques to other providers, such as offering clients support with basic needs during the initial outreach and hiring staff with lived experience.
Harris County, Texas Jail Diversion Center
Challenge: In Harris County, Texas, a Jail Diversion Center (JDC) offers food assistance, housing vouchers, and visits with doctors and mental health professionals instead of being booked into jail. Yet, many individuals brought in on misdemeanor charges left before finishing the initial assessment — often within an hour of arriving.
Data-driven Insight: Officials reviewed arrival and departure time stamps, as well as individuals’ progress across service milestones. Data showed the biggest problem was the long, repetitive intake process.
Initial Results: The JDC reduced the number of repeated questions and cut overall assessment length in half. Following these changes, 13% more JDC clients made it through the intake process. Initial evaluations showed the number of clients given a case manager — meaning they had stayed at least 24 hours — jumped by 80%.
Question 2: How long does each step take?
When clients need services, they do not want to wait. Delays in assessment, enrollment, or service provision can lead to a frustrating client experience at best and, at worst, drive clients away entirely. Reviewing data with a focus on time can help reveal spots where progress lags.
Michigan Health and Human Services
Challenge: In Michigan, Health and Human Services officials knew that many children who were removed from their biological parents and placed into foster care were not being connected to important mental health services at a critical point early on in the placement process.
Data-driven Insight: When officials looked into the timing of mental health referrals in two pilot counties, they learned only 25% of children received services during their first two months in care. Additionally, one in three children placed into foster care experienced a disruption, or change in placement, within 60 days — often the result of escalating behavioral challenges stemming from unmet mental health needs.
Initial Results: Officials implemented new protocols for trauma screening and referrals. Staff were advised to refer all eligible children to mental health services within two weeks of placement. Case workers also now discuss relevant trauma history and potential mental health needs at initial family team meetings and subsequent caseworker / service provider meetings. Now, half of removed children receive services within their first month of placement.
Florida Department of Children and Families
Challenge: In Florida, parental substance use was an underlying factor in the majority of child welfare investigations, yet many families were not able to access services after being referred.
Data-driven Insight: Officials at the Department of Children and Families reviewed service delivery data and found that only 7% of caregivers referred for substance use treatment received help within 30 days of a referral.
Initial Results:Providers responded by setting up telehealth appointments, creating priority treatment slots for referred parents, and sending text-message reminders. These changes and others reduced the time between assessment and referral by five days and doubled the number of families attending treatment within one month of a referral.
Rhode Island Department of Health
Challenge: In Rhode Island, Department of Health officials saw that some new parents had new-family visits scheduled too far into the future, which they knew resulted in frequent cancellations or missed appointments. Officials also knew that many supportive services had eligibility windows, like programs that only enrolled families with children younger than three months.
Data-driven Insight:To ensure this did not happen to families with serious needs, health officials reviewed vital records, prioritizing factors including low birth weight or medically fragile babies, as well as parents with a history of substance use.
Initial Results:A community health worker or nurse tried to connect with these families before they left the hospital in order to schedule the first appointment. These and other efforts were impactful at getting more families with serious needs to their first appointment.
Question 3: Who are we actually reaching?
Different people may experience the referral process differently. Some may face additional challenges in accessing or progressing through services. Breaking data down by race, ethnicity, gender, geography, or other priority population characteristics can highlight specific areas of success and struggle.
Chicago, Illinois Department of Family and Support Services
Challenge:Chicago’s Department of Family and Support Services (DFSS) wanted to expand their rental assistance program to those hit hardest by the COVID-19 pandemic: the city’s Black and Latinx neighborhoods.
Data-driven Insight: DFSS analyzed data on the race and ethnicity of individuals who reached different stages of the application process and learned that the share of applications from Latinx families was less than half of what they expected.
Initial Results: DFSS worked with trusted community messengers to share information about the rental assistance program to these specific audiences. As a result, 25% more Latinx families applied for rental assistance.
Memphis, Tennessee Shelby County School District
Challenge: In Memphis,Tennessee, Shelby County School District officials and community providers wanted to support chronically absent students, yet they lacked a consistent strategy for finding and enrolling those students most at risk of missing school.
Data-driven Insight: Officials used disaggregated administrative data on absences to identify whether chronically absent students were absent over long or short periods of time. They also ranked students by number of days missed.
Initial Results: This new data allowed providers to reach out to high-need families first. As a result, 17% more chronically absent students enrolled in support services across the test schools.
What next?
Congratulations! You started the process to identify the various steps in your referral process. You considered the three key questions:
What are the key referral steps and where are people dropping off?
How long does each step take?
Who are we actually reaching?
Now is the time to explore why gaps exist and what your jurisdiction can do about it.
Begin by looking for places where you can have the biggestimpact. For example, contacting 60% of clients — instead of 10% — is a far more significant improvement than just increasing the share of clients referred from 95% to 99%.
After all, using data to identify bottlenecks, backlogs, and breakdowns is not about getting perfect numbers — it is about making a positive impact in people’s lives.
Remember to start simple
It can be easy to get caught up in a quest to gather all the information, find the perfect data, or create the best organization system. Those efforts are less helpful than just picking a problematic spot in your referral process, focusing on what you have, and asking a different question. Conversations with clients and staff can reveal additional insights into challenges and opportunities going forward. Here is how two jurisdictions started simple:
Make existing data do more. In North Carolina, NCWorks Career Center staff created a new service code in their existing case management system to flag “long-term unemployed” clients. Staff could then easily pull reports for this priority population that focused on identification, service engagement, and employment outcomes.
Work with what you have. When the Chicago Department of Family and Support Services was piloting a new program, they thought about buying a new data collection tool. Instead, they used Excel to supplement their existing case management system. It was easy, low-cost, and allowed for iterations over time.
