The Impact of Behavioral Health Consultations on Medical Encounter Duration in Pediatric Primary Care: A Retrospective Match-Controlled Study

Abstract

Introduction:

The disproportionate time required to effectively manage psychosocial concerns is a key barrier to advancing delivery of behavioral care by primary care providers. Improved time efficiency is one potential benefit of the integration of behavioral health consultants (BHCs) into pediatric care, but few studies have systematically examined this outcome. We examined the impact of embedded BHCs on duration of medical encounters in a pediatric primary care clinic.

Methods:

We conducted a retrospective matched-pairs analysis of encounters involving behavioral consultations versus encounters for similar patients that did not include a consultation (N = 114) using electronic health record timestamp data. We examined both Medical Duration (i.e., medical provider services) and Total Duration (i.e., medical services + behavioral consultation).

Results:

Patient encounters involving behavioral consultation had a significantly longer (+11.23 min) Total Duration than matched controls, but significantly shorter (−11.67 min) Medical Duration.

Discussion:

The results indicate BHCs may improve primary care provider efficiency for patients with behavioral concerns, a notable finding given the impact of clinical time-constraints on important health care outcomes.

The time required to adequately manage psychosocial concerns is a significant barrier to behavioral services in pediatric primary care. Medical visits involving psychosocial concerns are significantly longer than those without (Meadows, Valleley, Haack, Thorson, & Evans, 2011), and primary care providers (PCPs) identify time-constraints as a key impediment to behavioral care (Horwitz et al., 2015).

Integration of behavioral health consultants (BHCs) into primary care holds potential for reducing PCPs’ time-burden whilst meeting behavioral health needs; however, few studies have attempted to objectively measure BHCs’ impact on clinical workflows. Gouge and colleagues (2014, 2016) found that BHC presence was associated with less PCP time for encounters involving behavioral and psychiatric concerns in a pediatric residency clinic and rural private practice. Those studies provide evidence that BHCs reduce the time-burden of PCPs; however, neither controlled for potentially confounding patient factors that are associated with duration of care (e.g., patient age; Norlin, Crawford, Bell, Sheng, & Stein, 2011), and both utilized direct observation methods, which may induce unintended observer effects. Investigations using methods to control for potential confounds are needed to determine if BHCs produce time-savings across different patient populations and clinical settings.

This study assessed the impact of behavioral consultations (we use “behavioral consultation” to refer to unscheduled patient-facing interactions provided the same day as medical care, throughout) on the duration of medical care in a pediatric residency clinic using retrospective electronic health record (EHR) data. We hypothesized that the duration of medical care for patients who received a behavioral consultation would be shorter than matched controls who did not receive a consultation.

Method

We conducted a retrospective chart review of all patient encounters from January to May 2016 in a resident teaching clinic within an urban academic medical center in the Pacific Northwest. We selected a period in the latter half of the training year, ensuring BHCs and residents were well-versed in the clinical model. Characteristics of the clinic and integration model are summarized in Table 1. Generally, BHCs were expected to balance direct clinical care with interdisciplinary consultation and educational efforts. Research procedures were approved by the Human Subjects Institutional Review Board at [redacted for blind review].

Table 1.

Clinic and integration model characteristics.

Characteristic	Description
Setting	Urban academic medical center
Patient population	Approximately 6,000 children and adolescents, estimated 50% publicly insured
Clinic staff	Attending physicians, pediatric residents, nurses, medical assistants, and social worker
Clinic schedule	Five weekday afternoons per week
Behavior health consultants (BHCs)	Two pre-doctoral clinical psychology interns supervised by licensed psychologists available for 2–3 clinics per week (.2-.25 full-time equivalent units total)
Integration model	The model approximated “Integrated Care” as defined by Blount (2003) and emphasized provision of brief, unscheduled, face-to-face visits with youth and families during both well-child and acute medical care. BHCs shared clinical and workroom space with medical providers
BHC activities	BHCs provided direct care, documented in the medical record, and engaged in education of pediatric residents, inter-professional consultation, and quality improvement initiatives.
Initiation of consultations	Behavioral consultations were not directly tied to systematic screening, but could have been initiated through requests by physicians, allied staff, or the BHC based on chart review or case-staffing.
Focus of consultations	Consults were typically problem oriented and consisted of brief assessment to determine the nature of the presenting problem and corresponding provision of recommendations, including reassurance, brief intervention, or referral to more intensive services. Based on review of BHC documentation, the most common presenting concerns were disruptive behavior (34%), emotional problems (23%), inattention and hyperactivity (13%), and toileting concerns (12%).

Measures

Total Duration and Medical Duration were calculated from electronic timestamps in the EHR. Timestamps were artifacts of clinical staff marking each phase of a patient visit (e.g., check-in, rooming, beginning PCP encounter, ending PCP encounter, check-out) in the EHR in real-time for workflow logistics purposes, in a manner similar to that described by Hribar et al. (2017). Total Duration was the duration from when the PCP began the encounter until the PCP encounter was marked complete, signaling readiness for immunizations or check-out paperwork. This generally included the PCP’s examination, staffing with an attending physician, and any counseling, consultations, or procedures (excepting immunizations). To calculate Medical Duration, we subtracted the duration of behavioral consultations from Total Duration. Behavioral consultation durations were extracted from the BHCs’ clinical documentation. Other relevant patient (e.g., age, sex, number of behavior problems), provider type (e.g., year of residency) and encounter (e.g., well-child or acute visit, scheduled duration) data were also extracted. Identification of behavioral problems was based on International Classification of Diseases, Tenth Revision (World Health Organization, 2016) codes in the patient’s chart, including both mental disorders (e.g., F33.9 “Major Depressive Disorder”) and codes capturing individual behavioral symptoms (e.g., R45.84 “Anhedonia”).

A research assistant manually extracted data from the EHR. The second author ([redacted]) duplicated extraction for 10% (n = 184) of cases to calculate reliability. Exact agreement was 98% across data types.

Outliers and Missing Data

Data from 1823 encounters were extracted from the EHR. Timestamp values were missing in 16% of encounters. To control for input-error (e.g., forgetting to mark an encounter complete) and resulting outliers, we used an outlier identification method (Hoaglin & Iglewicz, 1987) to determine likely erroneous values. An additional 8% of values were labeled as outliers and treated as missing, resulting in missing values for 25% of cases. We used a multivariate imputations by chained equations method to impute five datasets. Analyses were then conducted on all datasets. Herein we report the most conservative results (i.e., smallest magnitude effects) with respect to the primary hypothesis.

Analyses

We calculated descriptive statistics to characterize the sample, Pearson’s R to test for bivariate associations, and F-statistics to assess interactions between provider type and duration. We used a matched-pairs approach to test whether patient encounters with behavioral consultations had significantly shorter Medical Duration than similar encounters without consultations. Preliminary analyses indicated that scheduled visit duration, number of documented behavioral problems, and patient age, and provider type were all related to Medical Duration. Thus, we used the automated case-matching feature of SPSS (IBM, 2017) to identify control encounters matched on patient age (within 6 months), sex, number of behavioral problems, visit type, and scheduled duration. When possible, we also matched by provider type, because Gouge et al. (2014) found differing effects of BHCs on duration of care across different years of residency. Each matched pair consisted of two distinct patients. We conducted dependent samples t-tests to test between-groups mean differences.

Results

A BHC was present for 30 of 58 consecutive clinic days (52%), during which 61 behavioral consultations occurred. Consultations ranged from 5 to 60 min, with a mean duration of 23.36 min (SD = 10.94). Table 2 summarizes the characteristics of the sample and the matched groups. For all patient encounters, Medical Duration was significantly correlated with patient age, number of documented behavior problems, and scheduled duration.

Table 2.

Patient and encounter characteristics

	Entire Sample	Matched-pairs analysis
	(N = 1823)	Consultations (n = 57)	Controls (n = 57)
Patient age (years), M (SD)*1	4.7 (5.1)	7.2 (4.3)	7.2 (4.3)
Female, n (%)	928 (51%)	21 (37%)	21 (37%)
# of identified behavior problems, n (%)*2
None	1123 (62%)	20 (35%)	20 (35%)
One	390 (21%)	15 (26%)	15 (26%)
Two	165 (9%)	11 (19%)	11 (19%)
Three or more	145 (8%)	11 (19%)	11 (19%)
Visit type, n (%)
Well-child	845 (46%)	32 (56%)	32 (56%)
Acute	978 (54%)	25 (44%)	25 (44%)
Scheduled duration (min), M (SD)*3	33.6 (6.2)	34.9 (.49)	34.9 (.49)
Provider type, n (%)*4
1st year resident	750 (41%)	20 (35%)	20 (35%)
2nd year resident	522 (29%)	24 (42%)	23 (40%)
3rd year resident	427 (23%)	10 (18%)	11 (19%)
Chief Resident or Attending	124 (7%)	2 (4%)	2 (4%)

Note: There were no significant differences between the Consultations and Controls groups on any characteristic.

^*Denotes a significant correlation or interaction with Medical Duration in the entire sample.

¹r(1821) = .15, p < .001.

²r(1821) = .13, p < .001.

³r(1821) = .12, p < .001.

⁴F(1,1821) = 5.33, p < .001.

Controls matched on age, sex, visit type, scheduled duration, and number of behavior problems were identified for 57 of 61 behavioral consultations. All but one pair were also matched on provider type. Mean Total Duration (min) for encounters involving a behavioral consultation (M = 67.75, SD = 23.30) was significantly longer than controls (M = 56.61, SD = 24.48), t(56) = −2.52, p = .015. Mean Medical Duration for encounters with a behavioral consultation (M = 45.21, SD = 23.67) was significantly shorter than controls (M = 56.61, SD = 24.48), t(56) = 2.60, p = .012.

Discussion

In this retrospective matched-pairs study, patient encounters involving a behavioral consultation had a significantly shorter Medical Duration compared to controls matched on known drivers of longer appointments. Despite differences in setting and methods, our findings are similar to those of Gouge and colleagues (2014; 2016), furthering evidence that BHC presence improves the time-efficiency of medical care for patients with behavioral concerns. Potential benefits of time-efficiency should be weighed with the possible costs, including longer Total Durations and extended occupation of clinical space. Further, PCPs may only experience time-efficiency as beneficial if it allows for engagement in alternative valued activities, rather than increased demands, and patients may prefer extended interactions with their PCP to specialized consultation. Continued research is needed to evaluate stakeholders’ perceptions of the relative benefits and costs of BHCs.

This study demonstrates the potential utility of EHR data in the study of integrated primary care. As Hribar et al. (2017) highlight, timestamp data may be particularly useful for conducting large-scale investigations of clinic workflows when alternative methods (e.g., observation) would be too expensive or laborious. While this investigation used timestamp data from routine practice retrospectively, integrated primary care researchers should consider how these and other data (e.g., symptom measures) can prospectively be built into EHR workflows for easy extraction and analysis. This may be especially useful for making comparisons of interest in clinical settings where random assignment or elaborate data collection procedures are not feasible. In such cases, case-matching or cohort-matching designs can be used to partially mitigate threats to internal validity. Such methods are relatively common in medical research, but as of yet underutilized in the integrated primary care literature.

Limitations should be considered when interpreting this study. The training clinic setting may generalizability. For example, medical encounters were generally scheduled to be 30–40 min in duration, roughly twice what is allocated in private practice settings. The BHCs’ role involved considerable inter-professional education and a relatively low volume of patient contacts, likely in contrast with non-academic settings. While EHR timestamps appear to be a valid measure of workflows (Hribar et al., 2017), and we took steps to minimize the impact of missing and inaccurate data, the retrospective nature of the study limited our ability to validate timestamp values or control for biases at the time of input. The results likely overestimated the true durations of care, but there was no apparent systematic bias. Combining observational and EHR data to assess agreement between methods would strengthen future investigations. Our method of assessing behavioral complexity was also limited by reliance on available EHR data and should be considered a rough proxy. Finally, other patient characteristics, including medical complexity or language/cultural differences may also affect care duration. Future studies should assess and account for these potential confounds.

Overall, this study provides additional evidence for time-savings as a benefit of BHCs in pediatric primary care and illustrates the utility of retrospective EHR data for studying integrated care.