Meeting Home
Website Home

Breakout Session Discussion Summary: All children will be screened early and continuously for special health care needs

Harriette Fox, Fox Health Policy Consultants - Group Facilitator

Individually, many of the grantees described contributions their projects had made toward the goal of developing mechanisms to screen and identify children with special health care needs. As a group, the grantees have made a significant contribution to the ongoing project of field testing a variety of screening tools and strategies in a range of practice settings and types of communities. Examples of the grantees’ findings include:

• The New England SERVE project, through its collaboration with Neighborhood Health Plan, has developed an algorithm for identifying CSHCN in the plan’s administrative databases, using diagnosis codes, Medicaid rating categories, utilization records, prescription drug records, and information on the use of other state programs. The children identified through this analysis were then screened using the FACCT screener to compare the two methods. Of the children identified through administrative records, about 54 percent also met the FACCT screener, and 60 percent were reported (by their parents) to have a "special health care need." This project has also worked to develop other plan-based identification strategies, such as using triggers (including grievances and appeals) to add children to a master database. In addition, the plan has begun using the FACCT screener as the tool to assess the health of new members.

• In Northwest Washington, the health plan has used its administrative database to identify CSHCN using the NACHRI Clinical Risk Group (CRG) schema, which classifies children by the severity of their conditions. The results are then used to determine the intensity of case management interventions. This project has involved working with providers to improve their coding, and comparing the findings of the analysis to the results of the FACCT screen. In this case, the administrative data analysis identified 60 percent of the children identified through the FACCT screen, and tended to miss the less severe cases (although the CRG method was more likely to identify children with mental health conditions.) This project also examined the expenditures for services for the children identified, and found that ten percent of the children account for 50 percent of the expenditures, and that, for a pediatric population, outpatient pharmacy expenditures account for only 8 percent of costs.

• The Florida CMS program is required to identify children enrolled in Medicaid or SCHIP who are eligible for specialized services. To do so, the project has relied on one broad question on the programs’ application form, which results in a very high false positive rate and requires nurse case managers to devote extensive time to further screening and assessment. The project is now testing the use of the FACCT screen earlier in the process to see if that produces a smaller pool of potentially eligible children. So far, it appears that the screener does miss some CMS-eligible children, and is particularly likely to miss Hispanic children.

• The Hood Center project at Dartmouth is using ICD-9 codes on claims files to identify children who would benefit from care coordination. Although this method is vulnerable to coding problems, the project did identify 15 percent of its pool of 50,000 children in the HMO program, but found that a disadvantage of this method is that children could not be identified prospectively. For those children who have been identified and given case management services, the program has been able to demonstrate cost savings, particularly in inpatient and home health care services, a result that has been effective in attracting the attention and support of the plan’s administration.

• The project in Washingon State is also testing the FACCT screener, with results that are less impressive to the health plan, as it appears that the screener casts a wider net than anticipated. The project’s next step is to use the WeeFIM to assess the functional level of the identified children and to compare the results of the screener to the children’s claims histories (with parental permission).

• The Boston Medical Center project is comparing the results of the QuICCC screener with an analysis of diagnosis codes within an integrated delivery system in a high-risk inner-city community. The project has found that either method identifies approximately 30 to 35 percent of children, but that only 40 percent of these children are actually identified through both methods. Of the children identified, 40 percent have primary mental health diagnoses, and 60 percent have physical or developmental disorders. In conducting these assessments, the project found that response rates to a mailed survey were unacceptably low, but that the cost of in-person interviews in the clinics was prohibitively high.

• The Minnesota project is also comparing the results of the QuICCC and the CRG system. The QuICCC identified 37 percent of enrolled children as CSHCN, but when these children’s charts were reviewed, it was found that half had no special needs. The CRG method identified 17 percent of children, one-third of which had no special needs, according to a chart review. Few children were identified using both methods.

• The San Antonio ABC project is testing a two-level screening system, in which children are initially identified using a series of five questions on their Medicaid/SCHIP application form, followed by the FACCT screener for those children who appear to be eligible. It was reported that the participating plans were struggling with identifying children who were already enrolled.

The grantees identified a number of general challenges associated with the screening and identification of CSHCN using the methods being tested. These included the following:

• Systems based on the analysis of claims data depend on the quality and accuracy of physicians’ coding practices. In some cases, mis-coding may simply reflect the incentives of reimbursement systems; for example, physicians may code mental health problems as neurological problems so they can be reimbursed for services.

• Identification systems based on survey tools require a sufficient level of trust on the part of families if they are to describe their children’s needs honestly. Families may be reluctant to identify their children as having special needs for a range of reasons, both emotional (fear of labeling their child or holding him or her back) and legal or administrative (fear of identifying a pre-existing condition or alerting immigration authorities to their use of services). For families to respond accurately to surveys, the interviewer must make clear that the responses will be used in ways that will benefit the family and will not be otherwise disclosed.

• Cultural barriers can also hamper the effectiveness of survey instruments. The projects have found that in some minority communities, cultural differences in families’ expectations and responses to their children’s needs produce lower yields for survey instruments.

• For plans to fully support efforts to identify CSHCN, their administrators must be convinced of the benefits to the plan. Children identified must be linked to care coordination programs, and short-term savings (or other tangible benefits) must be evident.

• Finally, the operational structure of plans has been shown to be a barrier to effective implementation of identification strategies. Lack of coordination among plans member services, management information systems, and service delivery departments makes it difficult to share data efficiently.

Given these experiences, the group agreed on several major lessons:

• Different strategies are needed for different purposes. The experience of the grant projects has shown that each of the various strategies for identifying CSHCN has a particular strength and emphasis. Rather than attempting to select a single best method, the participants agreed that the focus should be on determining which methods are particularly well suited to specific purposes, populations, and needs.

• In promoting screening programs to families, "special needs"must be identified in functional terms, so that parents can link the definition to their children’s abilities and disabilities. Terminology based on needs may not be meaningful to all families, or it may not mean the same thing to all families.

• Similarly, terminology like "high risk" does not have the same meaning to a plan administrator that it may to a pediatrician or care coordinator. For plans, high risk is associated with high cost, and the population of children with high health care costs is a small subset of the population with complex needs.

• Finally, the use of a generic, categorical definition of CSHCN should be mirrored in the implementation of new models of pediatric practice that address generic, rather than disease-specific, needs.

###

End of Report

Return to Top of Page