Decreasing Nonurgent Emergency Department Utilization by Medicaid Children

METHODS

Medicaid-recipient children not currently enrolled in a managed-care program who sought care in the Children's Hospital Emergency Department, Columbus, Ohio, from January through August, 1991, for nonurgent problems were eligible for admission into the study. Many alternatives were available in the community for such care including evening and weekend office hours, after-hour telephone availability, free-standing urgent care centers, and other hospital EDs. Because all eligible children were in fee-for-service Medicaid, no gatekeeping barriers were set up to discourage nonurgent ED utilization. The nonurgent nature of their condition was determined by triage by a trained registered nurse on arrival in the ED as “stable, no noted distress; care could be delayed at least 4 to 6 hours and/or referred to a clinic for nonemergency treatment,” although no patient was turned away from the ED for care. Intervention workers were assigned to work at varying times and shifts in the ED to enroll patients. Patients were enrolled into the minimal intervention or case management group depending on which worker was on duty on that shift. Comparison group patients consisted of patients who did not meet an intervention worker because they presented at a time when no worker was on duty or at a time when the worker was busy speaking with other families.

Minimal intervention group families were enrolled by a clerical employee without specific health professional training. She carefully explained the value of preventive care and, in particular, the advantages of primary care for problems such as the concern bringing the family to the ED that day. If the family currently had a PCP with whom they were pleased, the worker emphasized the importance of continuity of care at that site. If the family did not currently have a PCP, the worker assisted the family in choosing a provider which seemed to meet their needs using the primary care referral book for reference. This referral book was developed through a mail and telephone survey of all pediatricians, family physicians, general practitioners, osteopathic physicians, and clinics in our county to determine the provider's willingness to receive referrals of Medicaid recipient children for primary care services. Additionally, information concerning hours/days of practice, the availability of public transportation and parking, the educational background of the physician(s), other services offered on site, and type of after-hours availability, if any, was collected. The worker used this referral book to assist the family in making the first appointment with the PCP and followed up to determine compliance. Workers for the minimal intervention group were involved with families only until the first primary care appointment was completed.

Enrollees into the case management intervention group were staffed by a case manager, either a pediatric nurse or pediatric social worker. An enrollment interview similar in content to that for the minimal intervention group was used except that more in-depth information concerning potential barriers to care was included by these health professionals. Again, the referral book was used to assist the family in identifying a PCP but in the case management group, the worker remained in contact with the family for up to 3 months. During this follow-up period, the worker provided case management and attempted to eliminate barriers that she or the family felt interfered with appropriate utilization of primary care services. The two case managers were assigned to varying days and shifts and approached families for enrollment as time allowed; therefore assignment to a nurse or social worker depended on which worker, if any, was available during the time the family presented for care. Families presenting to the ED during days/shifts not covered by an intervention worker and hence not receiving assistance in referral to a PCP formed the comparison group. To avoid inadvertent bias regarding day and/or time of ED presentation between intervention and comparison groups, workers were assigned to varying shifts and days of the week.

Outcome measures (dependent variables) for all three groups were based on claims paid data received directly from the Ohio Department of Human Services Medicaid Program billing computers and included all billings for services for enrollees during the 2 years after enrollment. These claims were grouped into four 6-month periods, that is, 0 to 6 months, 6 to 12 months, 12 to 18 months, and 18 to 24 months after the index ED visit.

Throughout the time period covered in this study, all welfare recipients had the option of voluntarily enrolling in one of two Medicaid health maintenance organizations (HMO)s. Changes from fee-for-service to managed care or vice versa could occur at the beginning of any month. Because claims data were not available for analysis during periods of HMO enrollment for any given child, or during periods when the child was not eligible for Medicaid coverage, days of fee-for-service eligibility were calculated for each child and compared among groups. All claims were classified into visit groups based on Current Procedural Terminology (CPT) codes (see Table1). The validity of our defined groupings of ED visits as urgent or nonurgent was determined by blinded independent review of medical records by two pediatricians for a random sample of 200 visits. Agreement was 97.6% between the two reviewers (κ-free coefficient⁶ = 0.951), 91.1% (κ-free coefficient = 0.823), and 91.8% (κ-free coefficient = 0.836), respectively, between the two reviewers and the CPT codes.

Primary care visits were classified as illness care visits and preventive visits based on CPT and International Classification of Diseases, 9th Revision (ICD-9) codes (Table 1). Total emergency visits includes urgent and nonurgent emergency care visits. All pharmaceutical claims were grouped together. Total health care expenditure included the above categories as well as all inpatient care, home care, ancillary services, and durable medical equipment.

The statistical analysis of the data included using the χ² test, the Kruskal-Wallis, and the Mann-Whitney statistics, the latter two tests reflecting the nonnormal distribution of health care utilization data. The study was approved by the Children's Hospital Human Subjects Review Committee.