A+KIDS, a Web-Based Antipsychotic Registry for North Carolina Youths: An Alternative to Prior Authorization

Characteristics of pediatric antipsychotic users as captured in the antipsychotic registry are shown in Table 1. A total of 5,532 unique patients were entered into the A+KIDS registry by prescribers. Eighty-one percent of patients were between the ages of seven and 12. Among this age group, 59% of antipsychotic users were non-Hispanic white and 29% of users were non-Hispanic black. More than 70% of pediatric antipsychotic users were male.

We examined a number of resource utilization features at the patient level. There were few apparent differences regarding resources being used by younger children versus older children. However, younger children were more likely than older children to have never used antipsychotics (24% versus 13%) and were less likely to have been taking the same antipsychotic for more than six months (38% versus 57%). A majority of prescribers seeing patients were psychiatrists, seeing 55% of the younger group and 63% of the older group. Further, most prescribers initiating antipsychotic treatment were outpatient psychiatrists; prescribers who were not psychiatrists initiated treatment with antipsychotic medication less often (28% and 19%) for patients ages six and under and for patients ages seven through 12, respectively. A small number of prescribers who were not psychiatrists indicated that patients saw a psychiatrist as well. A minority of patients saw other prescriber types, including nurse practitioners and physician assistant psychiatric specialty prescribers or developmental pediatricians.

Among patients ages six and under and those ages seven through 12, most were reported to be receiving psychotherapy. The most common types of psychotherapy among psychotherapy users were cognitive-behavioral or intensive in-home therapy. Prescribers did not know the type of psychotherapy being provided (22% and 23% of prescribers for younger and older children, respectively). The most common antipsychotic medications being used by patients in both age categories were risperidone, followed by aripiprazole and quetiapine. Few patients used other second-generation antipsychotic agents, first-generation antipsychotic agents, or long-acting injectable antipsychotics.

Table 2 describes the characteristics of prescribers using the A+KIDS registry. In the first six months of availability, the registry was used by 730 unique prescribers. A majority of these prescribers were physicians (73%), followed by nurse practitioners and nurses (15%) and physician assistants (8%). The mean number of patients entered into the registry per prescriber was 8.3, with a median of 2.0, whereas the mean number of unique registrations per prescriber (including duplicate registrations per patient for multiple medications) was 8.9, with a median of 3.0. Overall, there were 6,473 unique registrations, excluding duplicates from the same prescriber, for the same medication, or for the same patient. The top 10% of prescribers accounted for a majority (56%) of all antipsychotic registrations. The top 5% of prescribers accounted for 40% of registrations, with a mean of 66.4 registrations per provider in that subgroup over the six-month period.

Table 3 is a cross-tabulation of reported primary diagnostic indications and primary target symptoms for registrations. The top four diagnosis groups included mood disorders (predominantly accounted for by unspecified mood disorders, affecting 64% [N=1,261] of the mood group), autism spectrum disorders and developmental disabilities, disruptive behavior disorders, and attention-deficit hyperactivity disorder. Aggression, irritability, and impulsivity were the highest-ranking target symptoms across these top four diagnostic groupings with few exceptions. One notable exception includes the target symptom of psychosis for schizophrenia and psychosis (81%). Overall, however, psychosis as a target symptom accounted for 5% of reported target symptoms.

We linked prescription claims data to the registry data and examined the time between registry entry and an antipsychotic claim. Figure 1 shows that 50% of antipsychotic prescription claims (fills) were registered in the seven days before prescription fill and that 59% of prescriptions were filled within 14 days of registration. More than 90% of prescriptions were filled at some point after the date of the registry entry. Of note, 5% of faxed registry entries and 6% of online registry entries had a prescription claim preceding the date of the registration, with no claims after this date. In addition, 2% of online and 1% of faxed registry entries had no prescription fills in the claims data associated with information collected through the registry process. Upon matching antipsychotic claims to A+KIDS entries, it was noted that by the sixth month of the registry, 72% of antipsychotic claims in the Medicaid program were matched to an A+KIDS registry entry (Figure 2). A majority of these registrations were entered using the online A+KIDS system, whereas 19% were matched to a faxed entry. Figure 2 also demonstrates that total monthly antipsychotic fills after launch of A+KIDS were never as high as in any month in the prior year.

The North Carolina Division of Medical Assistance and CCNC implemented a Web-based registry to monitor the use of antipsychotic medications by children and adolescents in North Carolina. Within six months, most antipsychotic prescription fills were registered with A+KIDS and most prescriptions were filled within 14 days of registry entry. We used a relatively strict definition for registration for the purpose of analysis such that we may have artificially deflated registration rates. For example, if a child saw a prescriber and had the prescription filled the same day, before the prescriber was able to register the child, that fill would be counted as unregistered in the analysis. A very small minority of registrations were not matched to subsequent antipsychotic fills, implying that an antipsychotic medication was never filled or that a non-Medicaid source, such as private insurance, paid for the medication.

The relatively rapid prescriber uptake may be attributable to the large prelaunch prescriber education effort and to the campaign to engage other stakeholder groups such as mental health advocacy groups and the North Carolina Psychiatric Association. Ongoing outreach to providers was facilitated by a preexisting group of network pharmacists and psychiatrists associated with CCNC (17). Further, several features that separate A+KIDS from traditional prior-authorization programs may have improved participation: no restriction on the choice of medication, simple registration criteria, instantaneous approval, and interactive educational features. Future features, including more educational content, prescriber clinical and communication tools, and maintenance of certification credit hours tied to participation, may serve to enhance participation. Further enhancements will also allow providers to track patient information longitudinally and compare features of their patients and their prescribing with those of their peers. These enhancements are aimed to make the registry a more useful tool and to motivate compliance with prescribing and monitoring guidelines.

The initial findings presented in this report demonstrate the potential for a medication registry that captures clinical and utilization data not available from claims data alone. Several of the initial findings were provocative and raise significant questions about antipsychotic prescribing practice in North Carolina. For example, it is notable that a majority of registrations came from a minority of registered prescribers, suggesting highly varied practice styles or concentrated high-severity patient groups among certain prescribers or regions. Also, the data presented in Table 2 raise the possibility of differences in patients-per-provider and registration-per-provider rates between physicians and advanced practice medical professionals. The registry will need to mature before this question can be analyzed, but such future explorations of the role of subspecialty prescriber type and other practice-level factors in antipsychotic prescribing practice may help point the way to targeted provider education initiatives, targeted mental health resource allocation, and policy formation. There is also a preliminary suggestion that overall antipsychotic prescribing rates have dropped since A+KIDS implementation. A future, more detailed time-trend analysis is needed to verify this apparent trend, but if borne out, one could postulate a range of underlying mechanisms, from raised awareness regarding the risks of antipsychotics to the increased labor of registering that prescribers may now associate with prescribing this class of medications.

The clinical features of the group were notable in that aggression and irritability were among the top three target symptoms across nearly all diagnostic groupings. Unspecified mood disorder was the most common diagnostic group, which is notable because it is not a pull-down choice listed in the A+KIDS interface online and must be written in by prescribers. These findings are consistent with previous reports that the majority of antipsychotic prescribing to youths is off label. The validity of using off-label prescribing as a method for distinguishing non–evidence-based practice from evidence-based practice is often called into question by mental health clinicians. Of note, some psychotropic medications are used in the community and studied for years in open-label and small controlled trials before approval by the U.S. Food and Drug Administration is pursued or obtained. Given that off-label prescribing is so prevalent in child mental health and pediatric health care in general, this distinction may be far too blunt an instrument for determining approval in prior-authorization programs.

The findings reflected in Table 3 also demonstrate the difficulty of attaining diagnostic certainty in child mental health and the profound symptom overlap across childhood mental disorders. These findings may support the notion that symptom-based prescribing dominates over disorder-based specific treatment choices in child mental health. For example, aggression, although it may be a common associated feature, is not part of the DSM-IV criteria for any mood disorder, but aggression was overwhelmingly the top target symptom among children receiving an antipsychotic for a mood disorder.

A+KIDS currently targets children covered by North Carolina Medicaid, potentially limiting generalizability of these findings. As the registry matures, we will be able to evaluate how the registry population compares with the Medicaid population and the North Carolina general population. The clinical information gathered by A+KIDS is entered by prescribers, many of whom may not be trained mental health diagnosticians. Further, the possibility exists that busy prescribers may allow support staff to enter data on their behalf. By design, A+KIDS allows only primary diagnostic and symptom-based indications. This feature may not allow prescribers to illustrate the complexity or severity of their patient’s illness but was deemed necessary for streamlining of the prescriber interface. These limitations are inherent trade-offs for gathering point-of-care–based information. With Medicaid claims linkage, medication and dosing information as well as components of diagnostic data are verifiable. However, diagnostic data, symptom data, and some components of medication usage history data cannot be verified without a detailed chart review and independently performed direct diagnostic patient evaluation.

Medication adherence assessment is an inherent challenge in this large-scale initiative, but prescription claims data analysis techniques are able to reliably estimate adherence (19). The use of biologic metrics of adherence among subgroups of patients may further strengthen the findings of future work. Future analyses will require further linkage to claims data and will focus on several areas, including validation studies of prescriber-entered data, the evaluation of resource utilization and behavioral health consequences resulting from registry implementation, evaluation of polypharmacy rates and patterns, detailed evaluations of relationships between antipsychotic usage rates and prescriber subtypes, detailed descriptions and analysis of obesity rates among registrants, and the ongoing evaluation of the ability of the registry to influence prescribing trends over time.