The majority of individuals (63%) whose parents reported ESM at screening continued to show ESM ~4 weeks later. Persistent or increasing levels of ESM showed a strong association with BPSD diagnoses. Persistent ESM did not increase the odds of having other diagnoses or suicidal ideation/behavior. Persistently elevated PGBI-10M scores (≥20) appear to be a useful and fairly specific predictor of BPSD and not other diagnoses. However, only a minority of individuals with moderate levels of Persistent ESM+ met criteria for a BPSD diagnosis. Moderate levels of ESM also occur in individuals with other common disorders, such as ADHD.
Longitudinal assessment of manic symptoms is more helpful than a single assessment for predicting the presence of BPSD. The present findings support using two administrations of the PGBI-10M, even if only a brief period of time (approximately 1 month) elapses between assessments. Assessing stability over time in symptom level further enhanced prediction of BPSD diagnosis, despite the changeable and complex mood symptom patterns often seen in BPSD.1,8
Using a DLR approach increases the consistency of test result interpretation, improves accuracy over unaided interpretation, and reduces risk of overdiagnosing BPSD.50,51
In the DLR framework, combining results from two administrations appears quite useful for “ruling out” a BPSD diagnosis, even in clinical settings with a moderate base rate. Broader application of this approach may improve resource allocation (i.e., time, effort, cost).52
Adding a second PGBI-10M administration resulted in substantial improvement in detecting BPSD without inflating the false positive rate – avoiding the pitfall of “overdiagnosis.” Elevated scores that remain stable or scores that increase at follow-up should be viewed as a “red flag” requiring additional assessment.
The DLR framework may be enhanced by iteratively including family history. Existing evidence indicates a 5-fold (DLR=5) increase in the probability of BPSD when a first-degree relative is diagnosed with BPSD.10,53
Clinics that routinely use a broad-band instrument such as the Child Behavior Checklist54
might follow-up high scores on the Externalizing scale30
with a PGBI-10M, then repeat the PGBI before referring the family for a more detailed diagnostic interview that includes careful probing of BPSD symptoms. Using multiple gates would filter referrals and increase the procedure’s specificity.
The DLR approach is analogous to using a weather report. The report will sometimes be wrong, but it can be a guide for behavior. For example, if the report says 50% probability of rain, a reasonable response would be to bring a rain coat. Alternatively, if it says ~0% chance of rain, making plans to be outdoors would be appropriate. Adopting this system allows a person to make better choices over the long run but will not prevent all instances of getting rained on. In most assessment cases, a thorough clinical assessment ultimately will be required.
The simpler and more familiar approach involving averaging screening and baseline PGBI-10M scores resulted in only modest efficiency in detecting the presence of BPSD. This is to be expected in a cohort enriched for manic symptoms but not specifically ascertained for BPSD. The modest efficiency observed further supports a more nuanced approach – part of a broader clinical assessment strategy - that considers scores across two administrations.
Large increases (i.e., > 6 points) in PGBI-10M scores were rare in this cohort. A small group (n=11) of eleven individuals were ESM− at screening but progressed to ESM+ at baseline. Interestingly, these individuals showed a substantially higher percentage of BPSD diagnoses relative to individuals with consistently low scores (27.3% vs. 5.5%). The small group size precludes inferences, but future waves of follow-up may help to determine whether increases over time in PGBI-10M scores serve as a strong prognostic indicator of BPSD onset.
The LAMS cohort intentionally selected new outpatient children with high or low scores on the PGBI-10M. Thus, the present findings are particularly helpful for devising assessment strategies in outpatient settings. However, results may be less applicable to inpatient samples or the larger, non-clinical population. Furthermore, the variable time between screening and baseline assessments, while not altering results statistically, and the merging of BP-NOS with other bipolar disorders represent limitations that influence the generalizability of findings. Larger epidemiological studies will be needed to determine whether the present findings generalize to the non-clinical population.
Several important questions remain regarding the relationship between ESM and BPSD. Will youth with Persistent ESM+ without BPSD develop BPSD later? Will individuals with remitted ESM+ and BPSD show a rapidly fluctuating course of symptoms? How can repeated parent reports be combined with clinician observations or other risk factors to enhance detection of BPSD? Follow-up assessments of the LAMS cohort will be essential to providing answers to these questions. Empirical approaches, such as growth mixture modeling, are particularly promising for clarifying pediatric-specific BPSD phenotypes and developing clinically useful diagnostic classification.