Further examination of behavior during extinction-based treatment of pediatric food refusal

Christopher W. Engler, Vivian F. Ibañez, Kathryn M. Peterson, Ashley S. Andersen

First published: 15 August 2023

https://doi.org/10.1002/bin.1974Digital Object Identifier (DOI) view metrics

Sections

PDF

CITE

Tools

Share

Abstract

Researchers have demonstrated the effectiveness of extinction-based treatments to reduce challenging behavior. Although bursts and temporary increases in emotional responding may occur in some basic and applied studies, recent studies on their prevalence have shown that side effects are far from ubiquitous. Woods and Borrero (2019) determined that bursts of inappropriate mealtime behavior and increases in emotional responding occurred in 30% of data sets during extinction-based treatment of pediatric feeding difficulties. In the current study, we conducted a retrospective consecutive-controlled case series by evaluating treatment graphs of inappropriate mealtime behavior, negative vocalizations, active acceptance, and mouth clean for 60 children admitted to an intensive day-treatment feeding disorders program. We discuss the implications of these findings and provide general recommendations for the use of extinction-based treatment.

1 INTRODUCTION

Researchers use functional analysis methodology to identify the reinforcers that maintain challenging behavior (Beavers et al., 2013). After identifying the functional reinforcer, researchers could intervene to treat challenging behavior using any number of evidence-based strategies. Extinction, the termination of a reinforcement contingency that maintains a specific response, is widely used to treat challenging behavior and increase alternative behavior (e.g., Fisher et al., 1993; Forehand, 1973; Hagopian et al., 1998; Iwata et al., 1993). Researchers have used function-based extinction to reduce severe challenging behavior like aggression (e.g., Fisher et al., 1993), self-injury (Goh & Iwata, 1994), and elopement (Piazza et al., 1997). Extinction has also been used to increase appropriate behavior such as mands (Betz et al., 2013), functional communication (Fisher et al., 1993), variability in toy play (Lalli et al., 1994), and cooperation (Cote et al., 2005).

Extinction has substantial empirical support as treatment for inappropriate mealtime behavior (Bachmeyer, Piazza et al., 2009; Kerwin, 1999; Patel et al., 2002; Piazza, Patel, et al., 2003; Reed et al., 2004). Pediatric feeding disorders often develop as the result of multiple etiologies (e.g., underlying medical conditions, skill deficits, and change-resistance) and can lead to dangerous outcomes if left untreated. In many cases, children learn that eating can be associated with pain or discomfort, increased response effort, or unpredictable consequences and therefore begin to engage in challenging behavior during meals (e.g., head-turning, aggression, crying, and spitting food). Borrero et al. (2010) showed that the most common caregiver responses to child inappropriate mealtime behavior were removing the bite or drink or terminating the meal and providing attention (e.g., coaxing). Despite caregiver’s well-intentioned efforts to motivate their child to eat appropriately, the delivery of escape or attention may worsen behavior. In fact, results of functional analyses have shown that inappropriate mealtime behavior is often maintained by negative reinforcement in the form of escape from bites and drinks (Piazza, Fisher, et al., 2003; Saini et al., 2019).

Although extinction-based treatments are well supported, results of basic and applied research have demonstrated that extinction may be associated with a range of undesirable side effects. Side effects can include extinction-induced aggression, increases in emotional responding, and extinction bursts (Goh & Iwata, 1994; Lerman et al., 1999; Lerman & Iwata, 1995). Lerman and Iwata (1995) and Lerman et al. (1999) reviewed applications of extinction-based treatment in published studies on problem behavior and a day-treatment program’s records for self-injurious behavior, respectively. These researchers defined an extinction burst as an increase in responding during any of the first three treatment sessions above that observed in either the previous five baseline sessions or all baseline sessions if there were fewer than five. Results of these two studies showed that extinction bursts occurred in 24% and 39% of applications, respectively. In addition, Lerman et al. (1999) found there were increases in aggression in 22% of applications.

Woods and Borrero (2019) reviewed the treatment data for 10 children admitted to an intensive feeding disorders day-treatment program who received escape extinction as a primary treatment component. Using the same definition as Lerman and Iwata (1995) and Lerman et al. (1999), Woods and Borrero examined extinction bursts of inappropriate mealtime behavior and increases in emotional responding in the form of negative vocalizations. The authors found that both side effects occurred in 30% of applications. More specific, two participants displayed bursts of inappropriate mealtime behavior, two participants had increased crying relative to baseline, and one participant displayed both (i.e., burst in inappropriate mealtime behavior and increase in crying).

In a few recent studies, researchers suggested that practitioners use alternative interventions in lieu of escape extinction to avoid possible undesirable side effects. For example, Fernand et al. (2016) identified that antecedent-based treatments for pediatric feeding disorders are useful to avoid the negative side effects that may occur with the non-removal of the spoon. Others have reported that escape extinction is associated with decrements in treatment integrity due to bursting and aggression (e.g., Sira & Fryling, 2012). Although bursts and emotional responding occurred in only a third of participants from Woods and Borrero (2019), it is still important to continue investigating how often and the conditions in which undesirable side effects are more or less likely to occur. Practitioner concerns regarding escape extinction are valid given that extinction-based treatment may not always be possible or appropriate to maintain a child’s safety (Bachmeyer, 2009; Peterson & Ibañez, 2018).

To our knowledge, no other researchers have systematically examined bursts or other side effects of extinction-based treatment of inappropriate mealtime behavior. For the few studies that report side effects, there are limited details and no specific arrangements to evaluate side effects. That is, the authors have not provided specific definitions for bursts or other side effects, and the findings were secondary to the primary research question (e.g., Reed et al., 2004). Furthermore, no studies to date have systematically examined the emergence of appropriate behavior during initial extinction sessions relative to the emergence of challenging behavior. In addition, Woods and Borrero only included a small sample of participants (10 children), limiting the overall conclusions that can be generated.

In the current study, we replicated and extended Woods and Borrero (2019) with a larger sample size of children who received extinction-based treatment for their feeding disorders. We used the Lerman and Iwata (1995) definition to examine the prevalence of extinction bursts and levels of emotional responding for 60 children admitted to an intensive day-treatment pediatric feeding disorders program from 2006 to 2016. In addition, we expanded upon earlier studies by evaluating levels of active acceptance and mouth clean during those same treatment sessions. To our knowledge, no studies have examined and compared child responding across appropriate and undesirable topographies during the initial transition to extinction during the treatment of inappropriate mealtime behavior.

2 METHOD

2.1 Setting and materials

The authors conducted a retrospective consecutive-controlled case series (Hagopian, 2020) and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist. The authors started by reviewing the records of 154 children admitted to an intensive day-treatment pediatric feeding disorders program for the assessment and treatment of their feeding disorder between 2006 and 2016. The children attended the feeding program daily (Monday through Friday) between 8:30 a.m. and 5:00 p.m. for approximately eight to 16 weeks (depending on the severity of the child’s feeding difficulties, caregiver goals, and payer coverage). Therapists conducted all assessment and treatment sessions in 4-m by 4-m clinic rooms that were adjacent to a room with one-way observation and two-way communication. Participants sat in age- or weight-appropriate seating for feeding such as a highchair. The clinic room included food trays; gloves; a digital food scale; one or more tangible items (e.g., toy), if appropriate for the participant’s protocol; timers; and various feeding utensils (e.g., plastic bowls and small Maroon spoons). Caregivers selected the target foods from a range of food groups (e.g., proteins, grains, fruits, and vegetables) and a nutritionally and calorically appropriate liquid (e.g., Pediasure) that was recommended by the program’s dietician.

2.2 Inclusion criteria

Given that children were admitted for engaging in inappropriate mealtime behavior during meals involving solid foods and liquids, authors included data sets for treatments of both solids and liquids. Authors included children’s records if they met the following criteria. First, authors identified records with a functional analysis of inappropriate mealtime behavior relative to food refusal, liquid refusal, or both. Two master’s-level Board Certified Behavior Analysts® (BCBAs®), with at least 2 years of experience working with children with feeding disorders independently visually examined and categorized the outcome of each test and control comparison separately for the functional analysis graphs. The test conditions varied per participant, but options included escape, attention, and tangible. The authors calculated agreement by dividing the number of agreements (i.e., both agreed there was a function or both agreed there was no function) across test and control condition comparisons by the total number of comparisons and then converted the quotient to a percentage. The agreement between the two master’s-level clinicians was 98%. For all test and control comparisons that led to disagreements, two doctoral-level clinicians, with over 20 years of combined experience working with children with feeding disorders independently, visually examined and categorized the outcome (i.e., there was a function or there was no function) of each test and control comparison separately. The authors included data sets that identified at least one function (e.g., social positive in the form of attention or tangible and social negative in the form of escape) of inappropriate mealtime behavior for further analysis. The authors excluded all data sets that did not identify at least one function of inappropriate mealtime behavior (e.g., undifferentiated functional analyses) or for which the two doctoral-level clinicians disagreed. After applying these criteria, we included 181 data sets from 110 participants.

For the remaining participants, the first author identified records with a treatment of inappropriate mealtime behavior that included a baseline phase that was directly followed by an extinction phase. During baseline, the inappropriate mealtime behavior must have resulted in the previously identified functional reinforcer(s), and during the extinction phase, the functional reinforcer(s) were withheld following the first occurrence of inappropriate mealtime behavior. The first author only included the initial application of extinction-based treatment for food or liquid refusal that met the baseline and extinction phase criterion. Lastly, the first author excluded data sets for which there were not session notes, protocols, or operational definitions detailing the events of each treatment. Records from 60 participants met the above inclusion criteria, including 90 data sets (51 for the treatment of food refusal and 39 for the treatment of liquid refusal). See Appendix A for a flow chart depicting the inclusion and exclusion process.

2.3 Participants

Participants ranged in age from 8 months to 8.5 years (M = 3 years) and included 19 females and 41 males (see Table 1 for additional participant information). The authors reviewed caregiver-reported intake paperwork for medical and developmental histories and determined that the participants presented with varying medical histories, including 41 (68%) with nasogastric or gastrostomy tube dependence, 22 (37%) with a history of or medically managed gastroesophageal reflux disease, 22 (37%) born premature, and 26 (43%) with failure to thrive. Six (10%) participants were diagnosed with Down syndrome, and seven (12%) were diagnosed with autism spectrum disorder. A multidisciplinary team that regularly included a dietitian, pediatric gastroenterologist, master’s or bachelor’s level therapist with specialized training in pediatric feeding and behavior analysis, psychologist, and speech language pathologist with expertize in pediatric swallow safety, conducted an evaluation with each participant. During the evaluation, the multidisciplinary team confirmed the safety of oral feeding and medical stability of each child before the team recommended admission to the intensive day-treatment program. The multidisciplinary team only admitted a child who had severe food and/or liquid refusal or severe food selectivity with either previous outpatient therapy with limited to no progress, an emergent medical need, or both. Each multidisciplinary team member provided ongoing support and monitoring of participants through their admissions as needed.TABLE 1. Demographic information and participant history.

Participant	Sex	Age	Developmental history	Medical history
1	F	1	—	FTT
2	M	2	—	SBS, gastroschisis, TD, liver transplant
3	M	6	—	Low oral intake, TD
4	M	3	—	Low oral intake
5	F	8	Developmental delays	Syndromic diarrhea, colitis, TD
6	M	5	—	FTT, gastroschisis, CLD, PRE, dysphagia, multiple organ transplant
7	M	2	—	GERD, FTT, bronchopulmonary dysplasia, TD, PRE, dysphagia
8	M	3	—	GERD, FTT, low oral intake, TD, nephrogenic diabetes insipidus
9	M	2	—	Low oral intake
10	M	3	ASD	Low oral intake, anemia
11	F	1	—	FTT
12	M	4	—	GERD, CLD, SBS, TD, PRE, dysphagia
13	M	2	DS	FTT, PRE
14	M	1	—	FTT, malformation of the mandible
15	F	1	—	Congenital heart disease, TD
16	M	1	—	Respiratory syncytial virus, SBS, enteritis, TD, anemia, dysphagia
17	M	1	—	GERD, low oral intake, bronchopulmonary dysplasia, TD, PRE, hypothyroidism
18	F	1	—	FTT, CLD, SBS, necrotizing enterocolitis, TD, PRE, dysphagia, congenital hypothyroidism
19	M	1	—	GERD, FTT, eosinophilic esophagitis, TD, food allergies
20	M	6	ASD secondary to genetic anomaly	GERD, low oral intake, laryngeal cleft, CLD, gastric motility disorder, TD, PRE, dysphagia, hypotonia
21	F	8	—	GERD, FTT, low oral intake, TD, polymicrogyria, hypothyroidism
22	M	2	—	GERD, FTT, TD, PRE, dysphagia, chronic urinary reflux, chronic kidney disease
23	M	1	—	GERD, low oral intake, bronchopulmonary dysplasia, CLD, TD, PRE, dysphagia, hypoxema
24	F	2	DS	—
25	M	3	—	GERD, FTT, low oral intake, pulmonary atresia, TD, dysphagia, charge syndrome
26	F	3	Developmental delays	GERD, CLD, TD, PRE
27	F	3	—	GERD, FTT, TD, dysphagia
28	F	3	—	Inadequate intake, TD, liver transplant
29	F	<1	—	GERD, FTT, TD
30	F	1	—	GERD, FTT, TD
31	M	4	ASD, mixed receptive-expressive language disorder	Constipation
32	M	5	DS	Langerhans cell histiocytosis
33	M	1	—	FTT, low oral intake, aortic stenosis, TD
34	M	2	DS	Hypothyroidism
35	M	4	—	GERD, sensory processing disorder
36	M	5	—	FTT, SBS, TD, PRE, langerhans cell histiocytosis, multiple organ transplant, food intolerances
37	M	6	Global developmental delays, ASD, ADHD	Esophagitis, TD, PRE, dysphagia
38	F	1	—	Lung sequestration, SBS, mitral valve stenosis, TD, polysplenia
39	M	2	—	TD
40	F	1	—	Bronchopulmonary dysplasia, small muscular ventricular septal defect, TD, PRE, anemia, retinopathy of prematurity, dysphagia
41	M	3	ASD	Food allergies
42	M	1	—	GERD, FTT, CLD, TD, PRE, hemolysis, elevated liver enzyme levels, low platelet count
43	M	6	DS	Low oral intake, atrioventricular canal defect repair, TD, esophagitis
44	F	1	—	GERD, TD, PRE, dysphagia, food intolerances
45	M	1	—	Gastroschisis, transverse enteroplasty, TD
46	M	1	—	FTT
47	F	3	—	FTT, TD
48	M	3	—	GERD, PRE
49	M	2	—	GERD, FTT, CLD, TD, PRE
50	M	2	—	GERD, bronchopulmonary dysplasia, CLD, TD, PRE
51	M	<1	—	GERD, low oral intake, TD, PRE, hydrocephalus
52	M	2	—	Low oral intake, perihilar and peribronchial thickening with hyper-infiltrates, food allergies
53	M	1	—	GERD, FTT, TD, PRE
54	M	4	Developmental delays, ASD	FTT, reactive airway disease, TD, bilateral hydronephrosis, hearing disorder
55	F	2	DS	FTT, pulmonary hypertension, duodenal atresia repair, atrioventricular septal defect repair, TD
56	F	1	—	Necrotizing enterocolitis, SBS, TD
57	M	3	ASD	Low oral intake
58	M	3	—	Low oral intake, dysphagia
59	F	2	—	FTT, TD
60	M	1	—	GERD, FTT, vertebral abnormalities, CLD, tracheomalacia, subglottic stenosis, paraesophageal hernia, dextrocardia, TD, PRE, minor renal caliectasis

• Abbreviations: ADHD, attention-deficit/hyperactivity disorder; ASD, autism spectrum disorder; CLD, chronic lung disease; DS, Down syndrome; F, female; FTT, failure to thrive; GERD, gastroesophageal reflux disease; M, male; PRE, prematurity; SBS, short bowel syndrome; TD, tube dependency of any kind.

2.4 Therapists and observers

Therapists and observers (i.e., data collectors) were employees or trainees in the pediatric feeding disorders program and held or were in pursuit of an undergraduate or graduate degree in behavior analysis, psychology, or a related field. In addition, therapists and observers were certified in and trained to deliver CPR and basic life support. The program’s supervisors (e.g., BCBAs®, BCBA-Ds®, and licensed psychologists) who specialized in the assessment and treatment of pediatric feeding disorders taught therapists and observers to feed and collect data during the functional analyses, baseline, and treatment conditions.

Program supervisors used behavioral skills training including review of written protocols and operational definitions of primary dependent measures, observation of live sessions conducted by trained therapists, and role-play with immediate feedback (Mueller et al., 2003). Trainers provided therapists and observers with opportunities to ask clarifying questions during and after the review of the written protocols and operational definitions. Observers also collected training data with patients until they reached 80% reliability with previously trained data collectors across all dependent measures for three consecutive sessions. Therapists and observers continued to receive at least daily supervision from BCBAs®, BCBA-Ds®, and licensed psychologists during all admissions.

2.5 Dependent variables, response measurement, procedural integrity, and reliability

Observers collected data on laptop computers using DataPal 1.0 (e.g., Bullock et al., 2017) either in the session room or in the adjacent room through one-way observation. The authors examined inappropriate mealtime behavior, negative vocalizations, active acceptance, and mouth clean as the primary dependent variables for this study. Observers also collected data on correct utensil placement and incorrect attention as measures of therapist procedural integrity.

Observers scored each disruption or head turn as inappropriate mealtime behavior when the bite or drink was within arm’s reach of the participant and the participant used their hand to contact the utensil, food or drink, or the observer’s hand or arm anywhere from the elbow down while the therapist was presenting the bite or drink; threw food, liquids, or utensils; blocked their mouth with their hand, bib, or toys; or turned their head away from the utensil. Observers scored a head turn when the participant moved the middle of their mouth 45° or 6.3 cm in any direction except toward the utensil; then changed direction or paused for 1 s or more and then moved the mouth another 45° or 6.3 cm in any direction except toward the utensil. Observers did not record other challenging behavior (e.g., self-injury and aggression) in the definition for inappropriate mealtime behavior. Due to clinic changes over time, observers either depressed one key for each inappropriate mealtime behavior or depressed two separate keys, one for disruption and one for head turn. Observers also scored session time when a utensil moved within arm’s reach of the participant. Observers depressed a key that started a timer immediately meeting the above definition and immediately depressed the same key when the definition was not met. DataPal 1.0 converted the total frequency of inappropriate mealtime behavior per session to responses per minute by dividing the number of inappropriate mealtime behaviors during the session by the session time.

Observers scored negative vocalizations when the participant cried, screamed, whined, swore, or made negative or refusal statements about the mealtime context. Observers depressed a key that started a timer to score the behavior after 3 continuous seconds of meeting the above definition and depressed the same key that stopped the timer after 3 continuous seconds without the behavior. DataPal 1.0 converted negative vocalizations to a percentage by dividing the duration of negative vocalizations by the session duration and then multiplied by 100.

Observers scored active acceptance in two different ways depending on the level of independence, including the participant being fed by the therapist or the participant feeding themselves. If the participant was fed by the therapist, observers scored active acceptance when the participant opened their mouth and the bite or drink entered the mouth within 5 s of the initial bite or drink presentation. The therapist did not deposit the bite if the participant was engaging in negative vocalizations unless the participant leaned forward and opened their mouth. If the participant was feeding themselves, observers scored active acceptance when the participant deposited the bite or drink within 8 s of the initial bite or drink presentation. The percentage of active acceptance was calculated by dividing the number of active acceptances by the number of bite or drink presentations and multiplying by 100.

Observers scored mouth clean 15 or 30 s after the bite or drink entered the mouth if there was no food or liquid that remained inside the mouth (not due to expulsion or vomiting), with the exception of a pre-determined criteria, typically a pea-sized amount. Observers scored mouth clean if this definition was met regardless if the child had active acceptance or if the child accepted more than 5 s after the initial bite or drink presentation. The percentage of mouth clean was calculated by dividing the number of mouth cleans by the total number of bites or drinks that entered the mouth and multiplying by 100.

Due to clinic changes over time, observers scored correct utensil placement in one of two ways. Using the first method (94% of data sets), observers depressed a key to start a timer if the utensil was in the correct position according to the protocol and depressed the same key to stop the timer if the utensil was in the incorrect position according to the protocol (see below for protocol details). Observers did not depress the key to stop the timer after a participant expelled (i.e., spit out or allowed the food or liquid to fall out of the mouth, past the plane of the lips) unless 3 s had passed after the expel occurred and the utensil was not back in the correct position at that time. Using the second method (6% of data sets), observers depressed the key immediately to stop the timer when the utensil was greater than 2.5 cm from the correct position. That is, observers did press the key to stop the timer immediately after a participant expelled if the utensil was not back in the correct position at that time. For all data sets, DataPal 1.0 converted the duration of correct utensil placement to a percentage after dividing the duration of correct utensil placement by the session duration and then multiplied by 100. Observers scored correct utensil placement for an average of 99% (range, 63%–100%) of sessions across data sets. The mean correct utensil placement across data sets was 99% (range, 4%–100%).

Observers scored incorrect attention when the therapist provided attention (e.g., reprimands, coaxes, and gestures) that was not specified by the protocol, within 3 s of inappropriate mealtime behavior. The percentage of incorrect attention was calculated by dividing the occurrences of incorrect attention by the number of inappropriate mealtime behaviors in the session and multiplying by 100. Observers scored incorrect attention for an average of 94% (range, 0%–100%) of sessions across data sets. The mean incorrect attention was 0.01% (range, 0%–33%). Program supervisors re-trained therapists during or following sessions with low levels of integrity (e.g., provided immediate feedback and role-played).

A second observer simultaneously but independently scored inappropriate mealtime behavior, negative vocalizations, active acceptance, mouth clean, correct utensil placement, and incorrect attention for an average of 22% (range, 0%–75%) of sessions across data sets. The first author reviewed the included 90 data sets from the study to determine who served as observers (i.e., data collectors) and determined that 53 different observers collected data. The observers for data sets below 10% interobserver agreement (IOA) were the same observers for data sets above 10% with the exception of three individuals across three of the 90 data sets. With this information, along with high levels of IOA (see below) and the rigor of training (see above), the authors maintained all 90 data sets in the current study.

DataPal 1.0 separates the duration of each session into 10 s intervals. Observers calculated IOA for inappropriate mealtime behavior using exact agreement calculations and divided the sum of agreements (both observers scored the exact same number of responses across each 10 s interval) by the number of agreements plus disagreements (observers scored a different number of responses across a 10 s interval) and converted to a percentage. Observers calculated IOA for active acceptance, mouth clean, and incorrect attention using total agreement calculations by dividing the sum of agreements (both observers scored an occurrence or both observers did not score an occurrence across each 10 s interval) by the total number of agreements plus disagreements (one observer did and the other did not score an occurrence) and converted to a percentage. DataPal 1.0 calculated IOA for negative vocalizations and correct utensil placement using a partial-interval agreement method for which the lower amount of time was divided by the higher amount of time for each 10 s interval, averaged across intervals, and converted to a percentage. For example, if one observer scored the duration of negative vocalizations as 2 s and another observer scored the duration of negative vocalizations as 4 s in the same 10 s interval, the outcome was 50%. IOA across data sets for inappropriate mealtime behavior, head turns, and disruptions averaged 94% (range, 0%–100%), 97% (range, 43%–100%), and 97% (range, 10%–100%), respectively. The mean agreement values for negative vocalizations and correct utensil placement were 96% (range, 0%–100%) and 96% (range, 0%–100%), respectively. The average agreement values for active acceptance, mouth clean, and incorrect attention were 98% (range, 33%–100%), 98% (range 0%–100%), and 99.99% (range, 0%–100%), respectively. Program supervisors re-trained observers following two or more sessions with low levels of interobserver agreement.

2.6 General procedure

Each participant followed a schedule consisting of three to six meals per day that were each approximately 20–45 min. Meals included multiple sessions, each session consisting of five bites or five drink presentations, or a pre-determined time cap (e.g., 10 min), whichever occurred first. The number of sessions per meal depended on the duration of each meal and the duration of each session within the meal. The duration of a single session depended on the participant’s behavior. There were brief breaks between sessions and longer breaks between meals, during which participants engaged in leisure activities with therapists (e.g., play with toys and walks).

The start of each participant’s admission involved an assessment period so that therapists could create individualized treatment plans. The assessment period began with therapists asking caregivers to feed their child as they would at home. Therapists used these direct observations of caregiver-fed meals, along with additional caregiver report, to inform the conditions of functional analyses. Before the functional analysis, therapists conducted a paired-stimulus preference assessment (Fisher et al., 1992) to identify preferred tangible items (e.g., toys, iPad) to potentially use in treatment. Therapists conducted functional analyses of inappropriate mealtime behavior using procedures similar to Bachmeyer et al. (2009).

At the conclusion of the initial assessment period, a program supervisor met with at least one caregiver to discuss the outcomes of the assessments and provide the caregiver with a choice of treatment options, including an overview of the potential benefits and risks for different treatments. The program supervisor reminded the caregiver that they could stop treatment at any time. Caregivers were permitted and encouraged to observe all assessment and treatment sessions in the one-way observation room, and a program supervisor was always present with the caregiver(s) during transition sessions from baseline to treatment.

2.7 Data sets

2.7.1 Design

Therapists demonstrated functional control of the extinction-based treatment with a reversal design. The reversal designs sometimes included a multielement comparison evaluating other treatment components (e.g., re-presentation of expelled bites or drinks vs. no re-presentation). Although additional phases or treatment evaluations were included in some of the data sets for clinical purposes, the authors only analyzed the initial AB phases for the purpose of this study. The A phase consisted of baseline conditions based on the results of the functional analysis, and the B phase consisted of extinction conditions.

2.7.2 Baseline

During each session, the therapist presented the utensil to the participant either at the lips or on the table in front of the participant and said, “Take a bite (drink).” The therapist provided behavior-specific praise (e.g., “Great job taking your bite!”) if the participant actively accepted the bite or drink. Acceptance of the bite or drink initiated a pre-determined mouth-check interval that was typically 30 s in duration. The therapist conducted a mouth check when the interval elapsed and modeled an open mouth. The therapist provided behavior-specific praise for mouth clean. The therapist delivered a vocal instruction for the participant to, “Finish swallowing your bite (drink)” if food or liquid was in the participant’s mouth at the mouth check (i.e., packing). The therapist moved the utensil to the corner of the participant’s mouth during any coughing, gagging, or vomiting and then moved the utensil back to the center of the participant’s lips when the participant was no longer coughing, gagging, or vomiting.

The therapist followed the general functional analysis procedures and delivered the functional reinforcer(s) for either 15 or 30 s following the first occurrence of inappropriate mealtime behavior. For example, in an escape and attention baseline, therapists removed the bite and provided continuous attention in the form of coaxing following inappropriate mealtime behavior.

Additional components varied across participants. For a subset of data sets (7 of 90; 8%), therapists included additional reinforcement during baseline (e.g., differential reinforcement for active acceptance) beyond providing the functional reinforcer(s) for inappropriate mealtime behavior. For two data sets (2%), therapists implemented a multielement comparison of different reinforcement conditions (e.g., differential reinforcement for active acceptance vs. non-contingent reinforcement) while continuing to reinforce inappropriate mealtime behavior with the functional reinforcer(s). Other multielement comparisons included texture manipulations (1 of 90, or 1%; pureed vs. table-textured bites), independence (3 of 90, or 3%; participant feeding themselves vs. the therapist presenting bites or drinks), rate of presentation (1 of 90, or 1%; presenting bites every 15 s vs. every 30 s), bolus size (4 of 90, or 4%; e.g., 2-cc vs. 4-cc bolus), bolus consistency (1 of 90, or 1%; thick vs. thin liquid), utensil (5 of 90, or 6%; e.g., Nuk brush vs. small Maroon spoon), and re-presentation (1 of 90, or 1%; re-presentation vs. no re-presentation for expulsion).

2.7.3 Treatment

Therapists used the same procedures as baseline but no longer provided the functional reinforcer(s) following inappropriate mealtime behavior. During escape extinction with the therapist serving as the feeder, the therapist kept the utensil touching the participant’s lips until the participant opened their mouth, and the therapist deposited the bite or drink inside the mouth or until the predetermined session-time cap elapsed (e.g., 10 min). During escape extinction with the participant feeding themselves, the therapist placed the bite or drink on the table or tray in front of the participant and waited 8 s before implementing gentle hand-over-hand guidance to keep the utensil touching the participant’s lips. If applicable (across a subset of data sets), the therapist used the utensil to re-present expelled food or liquid by scooping up the food or liquid with the utensil, or gathering a fresh bite or drink and then kept the utensil touching the participant’s lips until the participant again opened their mouth and the therapist deposited the bite or drink. If the participant was expelling at the subsequent presentation interval, the therapist continued to re-present the bite or drink while instructing the participant to, “Swallow your bite (drink)” approximately every 30 s or until the bite or drink remained in the mouth for at least 3 s. Therapists never re-presented emesis (i.e., vomit) or contaminated food or liquid to the child. If food or liquid became contaminated (e.g., fell to the floor, mixed with mucous, and mixed with emesis), the therapist retrieved a fresh bite or drink for the re-presentation procedure. During attention extinction, therapists refrained from delivering attention following inappropriate mealtime behavior. If both functions were identified (attention and escape), therapists implemented both types of extinction at the onset of treatment.

Additional components varied across participants. For a subset of data sets (40 of 90; 44%), therapists included additional reinforcement-based components across all extinction sessions (e.g., differential reinforcement with an arbitrary tangible item for active acceptance) while no longer providing the functional reinforcer(s) following inappropriate mealtime behavior. For 25 data sets (28%), therapists implemented a multielement comparison of extinction with and without a reinforcement procedure. Other multielement comparisons included a texture comparison (1 of 90; 1%), independence comparisons (3 of 90; 3%), a presentation-rate comparison (1 of 90; 1%), bolus-size comparisons (4 of 90; 4%), a bolus-consistency comparison (1 of 90; 1%), utensil comparisons (6 of 90; 7%), and re-presentation comparisons (24 of 90; 27%).

For a subset of data sets (26 of 90; 29%), a second therapist provided assistance as a blocker (Wilkins et al., 2014; Woods & Borrero, 2019) to ensure participant safety and high levels of treatment integrity during some of the initial three extinction sessions. The trained blocker stood behind or next to the participant and placed their open hands above the participant’s forearms and upper arms, near the participant’s head to interrupt or block dangerous behavior (e.g., tipping the chair) or block attempts to push the utensil or feeder’s hand away. The second person never prevented the child’s movement or held the child (i.e., blocking was not a restraint procedure). The need to include the second person was continuously evaluated and monitored. As soon as therapists could implement the treatment with sustained high integrity and safety, the second person was removed. Observers scored child attempts to engage in inappropriate mealtime behavior when the second therapist was present (e.g., child pushing or hitting the second therapist’s hands or arms) and when the bite or drink was within arm’s reach of the participant.

2.8 Data analyses

The first and second authors independently visually examined each data set for the occurrence and nonoccurrence of an extinction burst of inappropriate mealtime behavior using Lerman and Iwata’s (1995) definition of an extinction burst as an, “…increase in responding during any of the first three treatment sessions above that observed during all of the last five baseline sessions (or all of baseline if it was briefer than five sessions)” (p. 93). The top and bottom panels of Figure 1 display representative data sets for the occurrence and nonoccurrence of an extinction burst, respectively. In the top panel, the highest rate of inappropriate mealtime behavior was 42.4 during the last five baseline sessions. The rate of inappropriate mealtime behavior increased to 85.2 during the first three sessions of extinction, demonstrating an extinction burst. In the bottom panel, the highest rate of inappropriate mealtime behavior was 78.3 during the last five baseline sessions. The rate of inappropriate mealtime behavior decreased to 27.8, 17.0, and 12.6 for the first three sessions of extinction, respectively, indicating no extinction burst.

The authors applied the same definition for negative vocalizations, active acceptance, and mouth clean to examine relative levels across the same baseline and treatment sessions. The top and bottom panels of Figure 2 display representative data sets for the occurrence and nonoccurrence of an increase in negative vocalizations, respectively. In the top panel, the percentage of negative vocalizations during baseline was stable at zero. The percentage of negative vocalizations increased to 39.5 during the first three sessions of extinction, demonstrating an increase in negative vocalizations. In the bottom panel, the percentage of negative vocalizations during the last five baseline sessions was zero. Negative vocalizations maintained at zero during the first three sessions of extinction, indicating no increase in negative vocalizations.

The top and bottom panels of Figure 3 display representative data sets for the occurrence and nonoccurrence of an increase in active acceptance, respectively. In the top panel, the percentage of active acceptance during baseline was stable at zero. The percentage of active acceptance increased to 80% during the first three sessions of extinction, demonstrating an increase in active acceptance. In the bottom panel, the percentage of active acceptance during the last five baseline sessions was zero. Active acceptance maintained at zero during the first three sessions of extinction, indicating no increase in active acceptance.

The top and bottom panels of Figure 4 display representative data sets for the occurrence and nonoccurrence of an increase in mouth clean, respectively. In the top panel, the percentage of mouth clean during baseline was at zero. The percentage of mouth clean increased to 100% during the first three sessions of extinction, demonstrating an increase in mouth clean. In the bottom panel, the percentage of mouth clean during the last five baseline sessions was zero. Mouth clean maintained at zero during the first three sessions of extinction, indicating no increase. It is important to note that the opportunity for mouth clean is dependent on the acceptance of food or liquid. That is, the participant could not swallow food or liquid if the food or liquid had not yet entered the mouth. For this analysis, we scored acceptance as zero for trials in which a child did not accept food or liquid. Therefore, we also scored mouth clean as zero during those same sessions.

When data sets included a multielement comparison of two treatments, the authors examined the data set by applying the definition for an extinction burst or increase in behavior, regardless of the changing conditions. For example, the first three treatment sessions could include two sessions of one condition (e.g., differential reinforcement and escape extinction) and one session of another condition (e.g., escape extinction).

The authors calculated IOA by dividing the sum of agreements (i.e., both authors scored occurrence or both authors scored nonoccurrence) by the sum of agreements plus disagreements (i.e., one author scored occurrence, the other author scored nonoccurrence) and multiplied by 100 for all dependent variables, for all data sets. The agreement between authors for inappropriate mealtime behavior, negative vocalizations, active acceptance, and mouth clean was 100%, 100%, 97%, and 96%, respectively. The third author independently visually examined each data set with a disagreement. All data sets were maintained and the agreement between authors is reported in the results.

3 RESULTS

Of the 90 data sets that met inclusion criteria, 51 (57%) were for the treatment of food refusal and 39 (43%) were for the treatment of liquid refusal. Inappropriate mealtime behavior was solely maintained by escape for 36 (40%) of the data sets and by escape and attention for 54 (60%) of the data sets. Figure 5 displays the results of the visual analysis of inappropriate mealtime behavior, negative vocalizations, active acceptance, and mouth clean across data sets. Six data sets (7%) met the definition for an extinction burst, 65 data sets (72%) met the definition for an increase in negative vocalizations, 40 data sets (44%) met the definition for an increase in active acceptance, and 63 (70%) met the definition for an increase in mouth clean. See Table 2 for the outcomes and treatment comparison descriptions.

TABLE 2. Outcomes and treatment descriptions.

P	Treatment context	Extinction burst	Increase in negative vocalizations	Increase in active acceptance	Increase in mouth clean	Function	BL comp	BL SR+	Tx comp	Tx SR+	Tx repres	Tx blocker
1	Solids	N	Y	Y	N	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	N	N	Y	Esc, attn	—	—	Repres	NCR	Comp	N
2	Liquids	N	Y	Y	Y	Esc, attn	—	—	—	—	Y	Y
3	Liquids	N	Y	N	Y	Esc, attn	Ind	—	Ind	—	Y	Y
4	Solids	N	Y	Y	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	Y	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
5	Solids	N	Y	Y	N	Esc	Utensil	—	—	—	Y	Y
6	Solids	Y	Y	N	N	Esc, attn	—	—	Repres	NCR	Comp	Y
	Liquids	N	Y	N	N	Esc, attn	—	—	SR+	No SR+ versus DRA versus NCR	Y	N
7	Solids	N	Y	Y	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	Y	N	N	Esc, attn	—	—	Repres	NCR	Comp	N
8	Liquids	N	Y	Y	Y	Esc	Bolus S	—	Bolus S	—	Y	N
9	Solids	N	Y	N	Y	Esc	Repres	—	Repres	NCR	Comp	N
	Liquids	N	Y	N	N	Esc, attn	—	—	SR+	No SR+ versus DRA versus NCR	Y	Y
10	Solids	N	Y	Y	N	Esc	—	—	SR+	No SR+ versus DRA	Y	N
	Liquids	N	Y	Y	N	Esc	—	—	Repres	DRA	Comp	N
11	Solids	N	Y	N	Y	Esc	—	NCR	—	NCR	Y	Y
	Liquids	N	Y	N	Y	Esc	—	—	—	NCR	Y	Y
12	Solids	Y	Y	N	Y	Esc, attn	—	—	Repres	—	Comp	N
13	Liquids	N	N	Y	N	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
14	Solids	N	N	Y	N	Esc, attn	—	DNRA, DRA	—	DNRA, DRA	Y	N
15	Solids	N	N	N	Y	Esc	—	—	—	NCR	Y	N
	Liquids	N	Y	N	Y	Esc	—	—	—	NCR	Y	N
16	Solids	N	N	Y	Y	Esc, attn	—	—	SR+	No SR+ versus DRA	Y	N
	Liquids	N	N	N	Y	Esc	—	—	Repres	DRA	Comp	N
17	Liquids	N	Y	Y	Y	Esc, attn	—	—	Repres	—	Comp	N
18	Solids	N	Y	Y	N	Esc, attn	—	—	—	—	Y	Y
	Liquids	N	Y	N	N	Esc, attn	—	—	—	—	Y	Y
19	Solids	N	Y	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	N	N	Y	Esc, attn	—	—	Repres	NCR	Comp	N
20	Liquids	N	N	N	N	Esc, attn	—	—	Repres	NCR	Comp	N
21	Solids	N	N	N	Y	Esc, attn	—	—	SR+	No SR+ versus DRA versus NCR	Y	N
	Liquids	N	Y	N	Y	Esc, attn	—	—	Repres	NCR	Comp	N
22	Solids	N	Y	N	N	Esc	—	—	SR+	No SR+ versus NCR	Y	N
23	Solids	N	Y	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
24	Solids	N	Y	N	Y	Esc	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	Y	N	Y	Esc	—	—	Repres	NCR	Comp	N
25	Solids	N	Y	Y	Y	Esc	—	—	Utensil	NCR	Y	N
	Liquids	N	Y	N	Y	Esc	—	—	Repres	NCR	Comp	N
26	Solids	N	N	Y	Y	Esc, attn	Utensil	—	Utensil	—	Y	Y
	Liquids	Y	N	Y	Y	Esc	—	—	—	DNRA, DRA	Y	N
27	Solids	N	Y	Y	Y	Esc	—	—	Repres	DRA	Comp	N
28	Solids	N	Y	Y	Y	Esc	—	—	Repres	DRA	Comp	N
29	Solids	N	N	Y	N	Esc	—	NCR	—	NCR	Y	N
	Liquids	N	Y	Y	Y	Esc	Bolus C	NCR	Bolus C	NCR	Y	Y
30	Solids	N	N	Y	N	Esc, attn	—	—	—	NCR	Y	N
	Liquids	N	Y	N	Y	Esc, attn	—	—	—	NCR	Y	N
31	Solids	N	Y	Y	Y	Esc, attn	—	—	Repres	NCR	Comp	N
32	Solids	N	Y	Y	Y	Esc	—	—	—	—	Y	N
	Liquids	N	N	Y	Y	Esc	Bolus S	—	Bolus S	—	Y	Y
33	Solids	N	N	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	N	Y	N	Esc, attn	—	—	Repres	NCR	Comp	N
34	Solids	N	Y	N	N	Esc, attn	—	—	Repres	NCR	Comp	N
35	Solids	Y	Y	N	N	Esc	—	NCR	—	NCR	N	N
36	Solids	N	N	Y	Y	Esc, attn	Ind	—	Ind	—	Y	N
37	Liquids	N	N	Y	Y	Esc	—	DRA	Repres	DRA	Comp	N
38	Solids	N	Y	N	Y	Esc, attn	—	—	—	NCR	Y	N
	Liquids	N	Y	Y	Y	Esc, attn	—	—	—	NCR	Y	Y
39	Solids	N	Y	N	N	Esc, attn	—	—	—	—	Y	Y
	Liquids	N	Y	N	Y	Esc, attn	—	—	—	—	Y	Y
40	Solids	N	Y	Y	Y	Esc	—	—	SR+	No SR+ versus NCR	Y	N
41	Solids	N	Y	N	Y	Esc	—	—	Repres	NCR	Comp	N
	Liquids	N	Y	N	Y	Esc	—	—	SR+	No SR+ versus DRA versus NCR	Y	N
42	Solids	N	Y	N	Y	Esc, attn	Text	—	Text	—	Y	Y
	Liquids	N	N	Y	Y	Esc, attn	Bolus S	—	Bolus S	—	Y	N
43	Solids	N	N	Y	Y	Esc	—	—	—	—	Y	Y
	Liquids	N	Y	N	N	Esc	Ind	—	Ind	—	Y	Y
44	Solids	Y	N	Y	N	Esc, attn	SR+	No SR+ versus DRA	SR+	No SR+ versus DRA	Y	N
	Liquids	N	Y	N	Y	Esc	—	DRA	Repres	DRA	Comp	N
45	Solids	N	Y	Y	Y	Esc, attn	Utensil	—	Utensil	—	Y	Y
46	Liquids	N	Y	N	Y	Esc	—	—	—	DNRA, DRA	Y	N
47	Solids	N	Y	Y	N	Esc	—	—	SR+	No SR+ versus NCR	Y	N
48	Solids	N	Y	Y	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
49	Solids	Y	Y	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	Y	N	Y	Esc, attn	—	—	Repres	NCR	Comp	N
50	Solids	N	Y	N	N	Esc, attn	Utensil	—	Utensil	NCR	Y	Y
	Liquids	N	Y	N	Y	Esc, attn	—	—	—	NCR	Y	Y
51	Solids	N	N	Y	Y	Esc, attn	—	—	—	NCR	Y	N
52	Solids	N	Y	N	N	Esc	—	—	Repres	—	Comp	N
53	Solids	N	Y	Y	Y	Esc	—	—	SR+	No SR+ versus NCR	Y	N
54	Solids	N	N	Y	Y	Esc	—	—	SR+	No SR+ versus DRA	Y	N
55	Solids	N	Y	N	Y	Esc, attn	Utensil	—	Utensil	—	Y	Y
	Liquids	N	Y	N	N	Esc, attn	Bolus S	—	Bolus S	—	Y	Y
56	Liquids	N	Y	N	Y	Esc, attn	—	—	—	—	Y	Y
57	Solids	N	Y	Y	Y	Esc, attn	Rate	—	Rate	—	Y	Y
	Liquids	N	Y	N	Y	Esc, attn	—	—	—	—	Y	Y
58	Solids	N	Y	N	Y	Esc	—	—	Utensil	NCR	Y	N
59	Solids	N	Y	N	Y	Esc, attn	—	—	SR+	No SR+ versus NCR	Y	N
	Liquids	N	N	N	Y	Esc, attn	—	—	Repres	NCR	Comp	N
60	Solids	N	N	Y	N	Esc, attn	SR+	No SR+ versus NCR	SR+	No SR+ versus NCR	N	N

• Abbreviations: Attn, attention; BL, baseline; Bolus C, bolus consistency; Bolus S, bolus size; Comp, comparison; DNRA, differential negative reinforcement of alternative behavior; DRA, differential reinforcement of alternative behavior; Esc, escape; Ind, independence; N, no; NCR, noncontingent reinforcement; P, participant; Rate, rate of presentation; Repres, representation; SR+, reinforcement; Text, texture; Tx, treatment; Y, yes.

Upon further analysis of undesired behavior, the authors determined that four data sets (4%) met the definition for both an extinction burst and an increase in negative vocalizations, two data sets (2%) only met the definition for an extinction burst, 61 data sets (68%) only met the definition for an increase in negative vocalizations, and 23 data sets (26%) did not meet the criteria for either measure. Extinction bursts and increases in negative vocalizations occurred more often for data sets with inappropriate mealtime behavior maintained by both escape and attention (i.e., four of six, 67% with extinction bursts and 37 of 65, 57% with increases in negative vocalizations) compared to escape alone (i.e., two of six, 33% with extinction bursts and 28 of 64, 44% with increases in negative vocalizations). Extinction bursts and increases in negative vocalizations occurred more often for data sets without a second therapist present to ensure participant safety and high treatment integrity (5 of 6, 83% and 42 of 65, 65%, respectively). In data sets without the use of the second therapist, 5 data sets (8%) met the definition for an extinction burst, and 42 data sets (66%) met the definition for an increase in negative vocalizations.

Upon further analysis of desired behavior, the authors determined that 27 (30%) met the definition for an increase in both active acceptance and mouth clean, 13 data sets (14%) only met the definition for an increase in active acceptance, 36 data sets (40%) only met the definition for an increase in mouth clean, and 14 (16%) did not meet the criteria for either measure. Increases in active acceptance and mouth clean occurred more often for data sets without a second therapist present to ensure participant safety and high integrity (30 of 40, 75% and 46 of 63, 68%). For data sets without the use of the second therapist, 30 (47%) met the definition for an increase in active acceptance, and 46 (72%) met the definition for an increase in mouth clean.

For desired and undesired behavior together, 54 data sets (60%) met the definition for at least one undesired behavior (extinction burst or increase in negative vocalizations) while also meeting the definition for at least one desired behavior (active acceptance or mouth clean). Only 13 data sets (14%) demonstrated an increase in undesired behavior (extinction bursts or negative vocalizations) in which there were no increases in desired behavior. Alternatively, 22 data sets (24%) demonstrated an increase in desired behavior (acceptance or mouth clean) in which there were no increases in undesired behavior. One data set (1%) did not meet the definition for an increase in either undesired or desired behavior.

There were 65 data sets with and 25 data sets without a multielement comparison during the first three extinction sessions. Extinction bursts occurred for 6% (4 of 65) and for 8% (2 of 25) of data sets with and without a multielement comparison, respectively. Increases in negative vocalizations occurred for 72% (47 of 65) and 72% (18 of 25) of data sets with and without a multielement comparison, respectively. Increases in active acceptance occurred for 45% (29 of 65) and 44% (11 of 25) of data sets with and without a multielement comparison, respectively. Increases in mouth clean occurred for 71% (46 of 65) and 68% (17 of 25) of data sets with and without multielement comparisons, respectively.

The authors analyzed the data with consideration to re-presentation as a treatment component during extinction including re-presentation for expulsion, no re-presentation for expulsion, and a multielement comparison between re-presentation and no re-presentation. A greater number of extinction bursts occurred without re-presentation (1 of 2; 50%) compared to with re-presentation (3 of 64; 5%). Only two data sets (8%) demonstrated extinction bursts during multielement comparisons of re-presentation versus no re-presentation. Increases in negative vocalizations occurred for 47 of 64 (73%), 1 of 2 (50%), and 17 of 24 (71%) of data sets with re-presentation, without re-presentation, and with a multielement comparison, respectively. Increases in active acceptance occurred for 32 of 64 (50%), 1 of 2 (50%), and 7 of 24 (29%) data sets with re-presentation, without re-presentation, and with a multielement comparison, respectively. Increases in mouth clean occurred for 46 of 64 (72%), 0 of 2 (0%), and 17 of 24 (71%) data sets with re-presentation, without re-presentation, and with a multielement comparison, respectively.

In addition, the authors analyzed the data with consideration of reinforcement procedures as a treatment component during extinction, including reinforcement for all sessions, no sessions, and a multielement comparison of reinforcement procedures. Extinction bursts occurred for 3 of 40 data sets (7.5%) with some type of reinforcement included, 1 of 25 data sets (4%) without reinforcement, and for 2 of 25 data sets (8%) with a multielement comparison of reinforcement interventions. Increases in negative vocalizations occurred for 27 of 40 (67.5%), 20 of 25 (80%), and 18 of 25 (72%) data sets with reinforcement, without reinforcement, and with a multielement comparison, respectively. Increases in active acceptance occurred for 14 of 40 (35%), 13 of 25 (52%), and 13 of 25 (25%) data sets with reinforcement, without reinforcement, and with a multielement comparison, respectively. Increases in mouth clean occurred for 29 of 40 (72.5%), 18 of 25 (72%), and 16 of 25 (64%) data sets with reinforcement, without reinforcement, and with a multielement comparison, respectively.

Finally, the authors analyzed data sets that included multielement comparisons of antecedent strategies. There were 16 comparisons of different antecedent strategies (e.g., thick vs. thin texture and smaller vs. larger bolus) during extinction-based treatment. Extinction bursts did not occur during any of the data sets in which there was an antecedent comparison. Increases in negative vocalizations occurred for 12 (75%), increases in active acceptance occurred for 9 (56%), and increases in mouth clean occurred for 13 (81%).

4 DISCUSSION

In the current study, we examined the prevalence of extinction bursts of inappropriate mealtime behavior, increases in emotional responding (i.e., negative vocalizations), increases in active acceptance, and increases in mouth clean (i.e., consumption) during initial exposure to extinction-based treatment of inappropriate mealtime behavior. For all participants, therapists first implemented functional analyses followed by a function-based treatment (i.e., escape extinction or escape and attention extinction) to reduce inappropriate mealtime behavior and increase appropriate alternative behavior, such as bite or drink acceptance. For a subset of cases, therapists also implemented reinforcement-based (e.g., differential reinforcement with arbitrary tangible items for active acceptance), extinction-based (e.g., re-presentation of expelled bites or drinks), or antecedent-based (e.g., texture manipulations) treatment components in addition to the reversal design to better understand the effects of these additional treatment components. Interestingly, the authors found that extinction bursts of inappropriate mealtime behavior occurred across few of the reviewed data sets (i.e., only six of 90), while increases in negative vocalizations from baseline occurred more often (i.e., 65 of 90). In addition, we found an increase in active acceptance (40 of 90) and mouth clean (63 of 90) across many of the data sets.

These outcomes are noteworthy and serve as an important replication and extension of the Woods and Borrero (2019) study, primarily because we observed a few critical similarities and differences. Overall, both studies confirmed that the side effects of extinction occur during the treatment of inappropriate mealtime behavior. Participants from the current study displayed fewer extinction bursts but greater increases in emotional responding relative to baseline than participants from Woods and Borrero. It could be that these differences are due, in part, to the larger sample size in the current study (i.e., 90 data sets relative to 10 data sets in Woods and Borrero). Alternatively, Woods and Borrero may have discovered different outcomes as a result of procedural differences in treatment implementation or data-collection measures across the clinical sites from which these data were collected and mined.

Given the larger number of participants who displayed an increase in negative vocalizations, researchers should continue to identify strategies that serve to mitigate the aversive qualities of the treatment or mealtime experience, or result in lower levels of negative vocalizations. Even though previous studies suggest negative vocalizations that occur at the onset of treatment are short-lived (e.g., Phipps et al., 2022), practitioners should evaluate levels continuously and consider alternative treatment approaches if negative vocalizations persist, including monitoring the child’s emotional responding and affect outside of meals and before and after clinical appointments (at home). In our daily clinical practice, we encounter many children who are happy and playful with therapists between sessions and meals, even during the early stages of treatment when negative vocalizations occurred during some meals. Future researchers should continue to collect data and assess child affect during behavior-analytic treatment (e.g., before and after meals and throughout the full treatment experience), the child and family’s quality of life (before, during, and after treatment), and other aspects of the child’s emotional state during treatment (e.g., indices of happiness and overall affect) to better understand the treatment effects.

This study extends Woods and Borrero (2019) given that we also analyzed the prevalence of increases in appropriate or desired feeding behavior. We found that both active acceptance and mouth clean increased in only three sessions for just under half (i.e., 44% increase in active acceptance) or nearly three-quarters of the data sets (i.e., 70% increase in mouth clean) included in this study. These immediate and meaningful changes in consumption are critical, especially for children with feeding disorders who may be at risk of invasive medical procedures like tube placement, growth failure, scurvy or nutritional deficiencies, severe, and chronic dehydration (Piazza & Carroll-Hernandez, 2004), and for those who have a history of previous therapy with no or insufficient progress toward eating and drinking. Increases in acceptance and consumption could create an opportunity for children to begin contacting the naturally appetitive properties of eating and drinking. These advancements can be life-changing for families who have encountered years of struggling to increase their child’s oral intake. Future researchers should continue to examine how quickly acceptance and consumption increase during various types of treatment.

Although bite or drink acceptance is an important behavior in the response chain of eating, it does not mean the child will immediately begin to consume (swallow) bites and drinks. For participants in the current study, we scored the trial as zero for mouth clean if acceptance did not occur (even though mouth clean was not possible when no food or liquid had previously entered the mouth). We also did not analyze how many participants accepted some bites or drinks during baseline and how occurrences of acceptance in baseline related to levels of mouth clean during initial exposure to treatment. Future researchers should examine these variables more closely and the conditions in which additional treatment components are necessary to increase one (e.g., mouth clean) beyond the other (e.g., active acceptance).

Although extinction is a highly effective and robust intervention for children suffering from feeding difficulties, researchers should continue to investigate its effects, both desirable and undesirable, and its long-term outcomes. It is important to note that 13 data sets (14%) demonstrated extinction bursts or increases in negative vocalizations for which there were no immediate increases in active acceptance or mouth clean. We did not continue to analyze those 13 data sets to determine when negative vocalizations or challenging behavior decreased, and when active acceptance and mouth clean increased, as those analyses were beyond the scope of this study, given our concentration on initial treatment exposure. However, future researchers should continue to identify when to transition away from extinction-based treatments if there is persistence of challenging behavior or negative vocalizations. As part of our regular clinical practice, we transition to alternative treatment strategies if we observe that negative vocalizations or inappropriate mealtime behavior persist for longer durations, or if caregivers request to change treatments. We recommend that clinicians continually monitor both desired and undesired behavior beyond the initial treatment sessions and have a plan for when to discontinue treatment if meals become unsafe or if there are prolonged negative vocalizations. Alternatively, 22 data sets (24%) demonstrated an increase in acceptance or mouth clean with no extinction bursts or increases in negative vocalizations. These data are promising in that they show that nearly a quarter of the cases produced the most optimal outcome (i.e., no increase in undesirable behavior while appropriate behavior increased). Data from this study also showed that 76 of 90 children demonstrated an increase in either active acceptance, mouth clean, or both, during the initial three treatment sessions.

Like Woods and Borrero (2019), some of our data sets included a multielement comparison. We cannot confirm whether rapidly alternating between treatments during comparisons differentially affected child responding. In addition, we did not analyze responding according to each condition of the multielement comparison so that we could retain our evaluation of the immediacy of extinction effects (i.e., increase during the first three treatment sessions relative to the preceding baseline sessions). There were more data sets with multielement comparisons of additional treatment components in this study (65 data sets) than without (25 data sets). However, the data reflected that extinction bursts or increases in negative vocalizations, as well as increases in appropriate behavior, occurred just as often when there was a comparison in place relative to when there was no comparison. Even though all comparison conditions included extinction-based treatment, this is a limitation of the current study and should be a focus of future research (e.g., use of other designs).

In terms of the additional components included across the multielement comparisons, the data sets included comparisons of reinforcement-based interventions, re-presentation (i.e., with or without, for expels), and of antecedent-based comparisons. The authors determined that there were a greater number of extinction bursts in treatments that did not include re-presentation (50%) relative to those that included re-presentation (5%). We also found that active acceptance (50%) was the same and that mouth clean (72%) increased more often during treatments that included re-presentation. However, also of importance, treatments including re-presentation resulted in a greater number of increases in negative vocalizations (73%). Clinicians should consider this outcome when making treatment selections and determine whether caregivers are interested in pursuing alternative options that result in similar increases in mouth clean but less negative vocalizations. Future researchers should also conduct larger-scale comparisons of treatments with and without re-presentation for expulsion to determine its effects on active acceptance and negative vocalizations.

Extinction bursts occurred regardless of whether some type of reinforcement was included relative to when there was no reinforcement. Interestingly, active acceptance increased more often across data sets without reinforcement-based components (52%) relative to data sets with reinforcement-based components (35%), suggesting that specific reinforcement may not be necessary to increase active acceptance in all cases. However, when reinforcement was included, there were fewer increases in negative vocalizations (67.5%) relative to when reinforcement was not included (80%). Regardless, more research is needed to continue comparing the difference in responding when extinction-based treatment includes or does not include reinforcement, especially given that relinquishing preferred items during differential reinforcement can pose challenges (Boyle et al., 2022) or caregivers may be reluctant to add unnecessary elements (Phipps et al., 2022). Of final note is that therapists always provided enthusiastic, behavior-specific praise following active acceptance and mouth clean, during all conditions and treatment variations across data sets. Therapists did not assess whether praise functioned as reinforcement for appropriate feeding behavior. In addition, we did not analyze whether praise had differential treatment or mitigation effects. Future researchers should continue to investigate the role of positive therapist attention delivered during treatment (e.g., praise for appropriate behavior and continuous casual interaction) to determine whether it could help to enhance treatment effects or reduce the aversive properties of the meal.

The initial moments after a previously reinforced response contacts extinction (no reinforcement) are critical to analyze to better understand the effects of extinction. Specific to pediatric feeding disorders, the only opportunity to score the primary target behavior for reduction (in this case, inappropriate mealtime behavior), is when a demand (e.g., a bite or drink) is present. This data collection structure requires that observers score the frequency of inappropriate mealtime behavior and the duration of time the demand is present to generate a rate of responding. Woods and Borrero (2019) calculated the rate of inappropriate mealtime behavior by dividing the frequency of inappropriate mealtime behavior by the time the utensil was within 2.5 cm of the child’s lips. However, the authors indicated that feeders held the utensil within 5 cm of the child’s lips during escape extinction. Therefore, it remains unclear whether observers from the Woods and Borrero study did not capture inappropriate mealtime behavior that occurred when the feeder followed the protocol of holding the utensil at a greater distance from the child’s lips. Future researchers should continue to explore how different parameters of the relevant establishing operations (e.g., spoon distance) or other aspects of data collection (e.g., operational definitions) for inappropriate mealtime behavior impact the assessment of extinction bursts.

Research has shown that extinction-based treatment of inappropriate mealtime behavior may be associated with the emergence of other side effects (e.g., extinction-induced variability). For example, Sevin et al. (2002) found that inappropriate mealtime behavior decreased, acceptance increased, and expulsion and packing emerged in the initial three sessions of extinction of inappropriate mealtime behavior. Woods and Borrero (2019) did not report data on the sequential emergence of other problematic or appropriate feeding behavior. Woods and Borrero reported an increase in acceptance across all participants, but it was not clear whether the increase in acceptance occurred during the initial three extinction sessions. Outcomes from the current study suggest that not only do desired behavior increase during initial sessions, but these increases occur for a majority of participants. Future analyses and research involving extinction bursts should continue to examine extinction-induced variability and the emergence of novel behavior to shed light on other potential meaningful direct or side effects of extinction.

To determine the prevalence of extinction bursts, we applied the same definition as Lerman and Iwata (1995) and Lerman et al. (1999), which involved measuring and comparing rate of inappropriate mealtime behavior between baseline and initial treatment sessions. The outcomes of this study are limited to that one dimension of an extinction burst (i.e., increase in rate). Only 7% of data sets displayed an extinction burst of inappropriate mealtime behavior, while 29% included the use of a second person for safety, which may suggest other dimensions are worth examining, including duration, magnitude, and latency of challenging or concerning behavior (Lerman & Iwata, 1996). However, in data sets without the use of the second therapist, only 8% met the definition for an extinction burst. These two percentages of bursts suggest there is also likely a range of variables that contribute to the likelihood of extinction bursts and increases in emotional responding in applied treatments that were not examined in this or other studies to date. It may also suggest that this subset of children engage in a range of potentially unsafe behavior during meals that may not precisely meet burst criteria or inappropriate mealtime behavior definitions.

Researchers have not agreed upon one precise definition for an extinction burst. Katz and Lattal (2021) referenced nine different extinction-burst definitions across various studies from 1995 to 2012. Along with Woods and Borrero (2019), we defined the extinction burst or increases in emotional responding, active acceptance, and mouth clean using the definition from Lerman and Iwata (1995). In an experimental analysis with pigeons, Katz and Lattal (2021) suggested that other methodological details should be considered, such as the level of response variability in baseline or the timing of transitions from baseline to extinction (e.g., between- or within-session shifts). Children with feeding disorders likely have unique reinforcement histories associated with eating. For example, some feeding difficulties develop when eating is associated with pain or discomfort (e.g., food allergies and reflux), and caregivers might respond to their child’s inappropriate mealtime behavior by providing reinforcement in the form of escape, attention, or a combination (Borrero et al., 2010; Piazza, Fisher et al., 2003). Although clinicians can estimate the length of time a behavior has contacted reinforcement by asking caregivers when the child’s feeding problems emerged, these learning histories are different from those that are well controlled in basic laboratories (St. Peter Pipkin & Vollmer, 2009). Given the high probability that treatment to increase acceptance and decrease inappropriate mealtime behavior will include escape extinction, the different conceptual frameworks from which to evaluate bursts should be applied to pediatric feeding disorders in future research.

Behavior analysts define a treatment as effective if there is a socially significant change in a socially meaningful dependent variable as a result of exposure to the independent variable (Baer et al., 1968). Yet, there are a number of other parameters that might affect a caregiver’s perception of treatment effectiveness and their choice of a given treatment. In some cases, the immediacy of the effect may be most critical, especially if the problem poses serious health risks or dangers to the child or others. For other caregivers, the absence of side effects may be most important. From an ethical standpoint, it is important that caregivers are informed of treatment options, informed of possible risks and benefits, and provided with choices (Beaulieu et al., 2018). This should include informing caregivers of potential risks like increased crying while also providing the potential benefits like increased consumption. In our daily practice, caregivers participate at every choice point and work with therapists to decide which treatment they prefer or which strategies they want to explore first (e.g., antecedent-, reinforcement-, and extinction-based components). As the healthcare provider, it is our responsibility to provide caregivers with education and data on all avenues and all potential effects. Given the severity of pediatric feeding disorders and that some children have experienced a long history of ineffective feeding therapies, caregivers may choose the potential risk of short-term increases in inappropriate mealtime behavior or emotional responding for the immediate, robust, and long-term benefits of the effective intervention. Phipps et al. (2022) demonstrated one such example that caregiver acceptability of extinction can remain high during periods of increased emotional responding. This may be in part due to the fact that the increase in unhappiness upon implementing extinction did not maintain for long durations and also corresponded with an immediate increase in acceptance for four of the six data sets. Results of the current study allow clinicians to better inform caregivers of the potential risks and benefits of extinction-based treatment. Additional research is necessary to assess the potential risks and benefits of alternative strategies and compare those with extinction-based treatments to provide caregivers with additional data-based options.

It is also often equally important for the treatment to be practical and durable across changes in the environment, such as setting (e.g., if meals occur at a family member’s home or at a restaurant); integrity errors or changes in feeder, given that multiple caregivers might feed a child across any given day (e.g., daycare provider, sibling, and nanny); or food or liquid differences across meals (e.g., changes in recipes, brands, and food preparation). Future researchers should continue to evaluate the short- and long-term effects of function-based extinction on the quality of life for children with feeding disorders and for their caregivers. In addition, researchers should include measures of treatment acceptability and preference and incorporate opportunities for child choice (e.g., Crowley et al., 2020) throughout treatment when possible.

In summary, a multitude of research has shown that extinction is an important and often medically necessary treatment component to reduce inappropriate mealtime behavior (e.g., Bachmeyer, Piazza et al., 2009; Kerwin, 1999; Patel et al., 2002; Piazza, Patel, et al., 2003; Reed et al., 2004). This study adds to the literature on extinction by examining the effects (desired and undesired) more closely, providing researchers, clinicians, and ultimately caregivers, more information on what to expect during extinction. One challenge that remains for clinicians is determining their competency (e.g., amount of experience, training, and education) to treat children with feeding disorders (Tereshko et al., 2021). Although our study showed immediate and robust effects of extinction and low levels of bursts, we still observed a relatively high number of data sets with increased negative vocalizations; and there are many other situations in which extinction might not be appropriate. Escape extinction and some procedures to promote swallowing (e.g., Nuk presentation; Milnes et al., 2019) require specific expertize, particularly when we consider how the esophagus and trachea (airway) are positioned next to one another anatomically, which increases the risk of aspiration. Of concern is that not all clinicians will know how to detect safety concerns or have access to other healthcare professionals who possess the necessary expertize for ongoing consultation (e.g., experts in pediatric swallow safety with specialized training). Therefore, we cannot overstate the importance of interpreting the findings of the current study with caution and encourage clinicians to practice within their scope of competence.

By gaining a clearer, data-based understanding of the potential treatment side effects and examining their prevalence during function-based extinction, clinicians will be better equipped to predict and mitigate undesired side effects while increasing beneficial outcomes. However, it is also important to identify the variables that are most critical for the child and the child’s caregivers when intervening on challenging behavior that could be life-threatening if left untreated or if not treated immediately.

ACKNOWLEDGMENTS

Thank you to the Munroe-Meyer Institute’s Pediatric Feeding Disorders Program staff for their assistance with this study.

1.  

CONFLICT OF INTEREST STATEMENT

The authors disclose no conflict of interest.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.