Background:

Children with cerebral palsy (CP) often receive outpatient physical therapy (PT) services, although the dosing schedule can vary widely depending on the child and the setting. Parents could make more informed choices about their child's therapy schedule if they knew whether a high-intensity (HIP) schedule of therapy was better than or equivalent to a usual model of weekly therapy.

Objectives:

The main objective of the ACHIEVE study was to (1) compare the short-term and long-term effects of 2 treatment service delivery models: 1 hour per day, 1× per week for 40 weeks (weekly); and 2 hours every weekday for 2 periods of 10 consecutive weekdays (total of 4 weeks) of a repeated periodic (HIP) bout of outpatient PT. We hypothesized that HIP therapy would result in comparatively greater improvements than would weekly therapy. In addition, we aimed to (2) determine individual differences in (a) children's response to treatment and (b) parent preferences; (3) monitor the progress and variability of motor function; (4) explore patient and family factors associated with treatment response; and (5) identify relationships between the assessment of the child's function by parents and the child's gross motor function as evaluated by professionals.

Methods:

We enrolled 100 children aged 2 to 8 years with CP in our study, which took place over a 9-month period. Forty-six participants were randomly assigned to either the weekly or HIP treatment group. Fifty-four participants self-selected their treatment group; this option was added to the study for families that could not commit to accepting either of the schedules. Trained therapists assigned each child a Gross Motor Function Classification System (GMFCS) level and completed baseline and 9-month testing. Testing included the Gross Motor Function Measure-66 (GMFM-66) score (primary outcome measure), as well as secondary measures, including Goal Attainment Scale (GAS) and Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III) scores. Parents also completed questionnaires on their children's skills, such as the Pediatric Evaluation of Disability Inventory-Computer Adaptive Test (PEDI-CAT). We analyzed the participants' scores using a linear mixed model, comparing the weekly and HIP groups in both the randomized and self-selected groups.

Results:

Both the weekly and HIP groups improved from baseline to 9 months in both the randomized and self-selected cohorts. At the time of this report, there was a 76% follow-up rate for the primary outcome measure at 9 months. There were no significant differences between the HIP and weekly groups for either cohort on the GMFM-66 (−0.92; 95% CI, −2.83 to 0.98; P = .34). There were also no significant differences between groups on the GAS, Bayley-III, or PEDI-CAT, although all groups improved on these measures over time, with significant main effects for time and clinically meaningful differences. Children with lower GMFCS levels (better function) and higher parent satisfaction with treatment made greater gains in motor function.

Conclusions:

HIP and weekly outpatient PT treatments, when provided using the same total number of hours, are likely to produce similar motor outcomes for children aged 2 to 8 years with CP. Parents and therapists can decide which schedule will be more beneficial for an individual child based on their unique personal and family needs.

Limitations:

A limitation of this study was a possible lack of power due to attrition during follow-up, which might have limited our ability to detect differences between groups and generalize the results. We may not be able to tell which changes may be due to normal development rather than the interventions, because the interventions were only 9 months long.

Impact of the Condition on the Health of Individuals and Populations

CP is the most prevalent childhood motor disability. In 2013, there were ∼350 000 children in the United States between 2 and 8 years of age with CP (the same range as the patients in this project).^4-7 CP affects 1 in every 474 children, and at least 10 000 children are newly diagnosed with CP each year in the United States alone.^2,8 CP is a nonprogressive disorder of tone, posture, and movement that causes motor disablity.^9,10 In most cases, CP is the result of a brain lesion that happens in infancy, meaning that there was an injury to the brain, which neither worsens nor improves. Infants born preterm are at high risk for brain injuries due to immaturity of the brain and lungs. Importantly, rates of preterm birth are rising, potentially resulting in more and more children with CP. Every week, 1509 infants are born at <32 weeks' gestational age.^7,11,12 This means that 1509 infants every week are born at least 8 weeks early and are at very high risk for CP. The average age of CP diagnosis is 19 months of age, and diagnosis is improving with early diagnostic criteria.^13,14

Children with CP are classified in Gross Motor Function Classification System (GMFCS) levels I to V from least to most severe motor involvement. Children in level I demonstrate the least impairment but often have trouble with high-level motor skill activities, such as playing sports or having precise coordination. Children in level V have very serious impairments and have no (or very limited) voluntary movement control. Overall, infants and children at all GMFCS levels show delays in motor development. Even small improvements in motor skill development during childhood have the potential to compound with growth and development, propelling children into new developmental trajectories that improve their health care, quality of life (QOL), reliance on caregivers, and ability to enjoy the basic human right to move.^15-20 In previous studies of intensity of rehabilitation, patients at higher GMFCS levels have typically been excluded.

Most children with CP have a life expectancy that mimics that of the general population, meaning that health care for patients with CP is lifelong and comprehensive.^9,21 The yearly economic cost associated with CP in the United States is estimated to be at least $11.5 billion.⁵ There are far-reaching consequences of CP, including health, functional status, and social participation that impact the child (patient), family, and society.¹⁰ CP is chronic and is associated with multiple other chronic conditions that require lifelong health care. In addition to gross- and fine-motor disabilities, there are co-occurrences of learning, cognitive, and social/emotional disability in children with CP.²² Children with CP often have other impairments in cognition, language, engagement, and participation in play. Secondary impairments from abnormalities in tone, posture, and movement are common and affect bone, joint, muscle, and cardiovascular health.^23,24 Behavioral disorders and epilepsy are also common comorbidities.^25-29 Most, if not all, patients with CP have more than 1 chronic condition,^24,30,31 further limiting their independence and participation.²⁴ CP can pose a major challenge for the independence and health of individuals.

CP makes up 20% to 25% of all childhood motor disabilities.^4,32 Outpatient physical therapy (PT) is a main service for children with motor disabilities and CP.^33-35 As such, in a typical day, a pediatric therapist may spend at least 20% to 25% of their time providing treatment to children with CP. The intensity of pediatric rehabilitation and therapy services for children with CP has been identified as a priority by national organizations and parent groups.^36-39 Establishing intensity parameters for outpatient pediatric PT treatment is critical for informing clinical decision-making, health policy, and guidelines for reimbursement. The intensity and timing of therapy interventions have implications for shaping the future of pediatric practice and improving motor and developmental outcomes for children with CP.

The ACHIEVE study was a comparison of the effectiveness of 2 intensities of treatment for children with CP in an outpatient PT setting. High-intensity periodic (HIP) therapy is 2 bursts of daily and high-intensity treatment. In comparison, a weekly schedule spreads out the therapy over time. Usual weekly therapy is considered the standard of care, although high-intensity models are also used clinically for children with CP. Importantly, the total number of hours of prescribed therapy was the same for the 2 groups, allowing for a powerful and scientifically credible comparative effectiveness study because only the distribution of the hours differed in the compared interventions. Most children with CP receive usual weekly therapy treatment; however, some do receive bouts of intensive therapy. Families may prefer one schedule over the other for a variety of reasons, including personal schedules, child temperament, and current therapy goals and priorities. The ACHIEVE study helped answer the question, “What are my options for intensity of therapy, and what are the benefits and harms of higher intensity and usual weekly programs?” In addition, it answered the question, “Given my child with CP's personal characteristics, conditions, and preferences, what changes should I expect in my child if I enroll in a high-intensity or usual weekly therapy program?”

Gaps in Evidence

There are wide variations in the number of hours of PT provided to a patient per month in current practice, suggesting clinical uncertainty.³⁵ The Institute of Medicine's top 100 priorities for comparative effectiveness research identify a comparison of “the effectiveness of focused intense periodic therapy and usual weekly therapy in managing cerebral palsy in children”³⁸ as a gap in knowledge. Two recent manuscripts summarized the results of a think-tank-style conference on dosing of therapy services for children with CP. Parent panelists, adults with CP, and scientists contributed to the conference. In summary, they stated:

Dosing of pediatric rehabilitation services for children with cerebral palsy (CP) has been identified as a national priority. Establishing dosing parameters for pediatric physical therapy interventions is critical for informing clinical decision making, health policy, and guidelines for reimbursement.³⁶

And:

There is no consensus on what the basic parameters should be for different treatment protocols. A very important parameter of intervention that is pivotal for treatment efficacy is dosing.³⁹

HIP and usual weekly dosing models are currently used in clinical practice to provide therapy for patients with CP. The most common option for intensity of therapy is weekly therapy for 30 minutes to 3 hours per week or even per month.^33,40 For children with CP, weekly therapy can continue for most of their childhood, with breaks in services during changes in patient and family priorities, such as increased participation in extracurricular or community activities. HIP programs, which typically provide therapy for several hours per day for several days in a row and are then repeated months later, are also available options. However, access to treatment is a potential barrier to this type of service delivery for several reasons, including cost, availability of this type of program, and family schedules. Importantly, high-intensity models have used specific types of treatment, such as treadmill training or constraint-induced movement therapy (CIMT), and have demonstrated rapid improvements in motor development, suggesting they could be more effective than usual weekly therapy.^15,41-44 There are 2 issues of immediate importance. First, until now, no one has compared the service delivery of usual weekly treatment with HIP treatment in an outpatient setting. Second, none of the high-intensity treatments in previous studies were compared with equal total hours of usual care, making it impossible to know whether the type of treatment, total amount of treatment, or intensity of treatment is the key ingredient.^36,39,45 The ACHIEVE study filled this critical gap by using 1 treatment type (ie, outpatient pediatric PT based on principles of motor learning, which was standardized and examined for treatment fidelity)) and comparing 2 intensities (HIP and weekly) in conducting a randomized controlled trial (RCT) for outpatient therapy treatment of children with CP, allowing for findings to be easily and immediately disseminated and translated into current practice. Table 1 describes the dosing and frequency terminology used in this report.

Table 1

Dosing and Frequency Terminology Used in This Report.

Potential for the Study to Improve Health Care and Outcomes

CP is primarily a disorder of posture and movement. Many children with CP, especially young children, withdraw from physical exploration.^46-51 In turn, this limits the exploration of more difficult and complex motor skills and the refinement of newly learned motor skills for improved coordination, play, and social engagement. As a result of delayed motor skills and decreased opportunity to physically explore, development across domains, such as language and cognition, is hindered.^49,51,52 Improving development, independence, productivity, health, and QOL may prevent major lifelong impairment and decrease caregiver burden associated with CP; both are of high significance.^5,53

The ACHIEVE study compared 2 intervention strategies delivered in real-word settings. In this study, the type of treatment was outpatient PT. PT can be delivered individually or in groups and can be provided in the home, in a clinic, in a school, and at other community locations. Outpatient PT is defined in this report as one-on-one skilled therapy with a therapist and a child in an outpatient clinic. Therapy approaches based on principles of motor learning,^54-56 functional training,^15,57-59 and goal setting^60-62 are the most commonly used and effective approaches for improving motor development in young children with CP.⁶³ Motor learning, functional training, and goal setting are the guiding principles for each treatment session in our clinics and for the ACHIEVE study. The aims of our research are consistent with dosing and intervention needs identified as important by patients, caregivers, clinicians, and research groups.^36,38,64 Our specific aims relate to the significance of our project by determining the immediate and short-term effects of usual weekly and HIP therapy, determining individual child characteristics that mediate benefits or harms of each intensity model, and beginning to determine the longer-term effects of usual weekly and HIP therapy.

How the Research is Focused on Questions That Affect Outcomes of Interest to Patients and Their Caregivers

Comparing focused HIP therapy and usual weekly therapy in managing CP in children is one of the Institute of Medicine's top 100 initial priority topics for comparative effectiveness research .³⁸ Pediatric physical therapists were surveyed in 2011 on the area of greatest need for PT research. “Dosing of therapy for children with CP” was identified as the topic with the largest need.²² As such, the Pediatrics Section of the American Physical Therapy Association and the National Institute of Child Health and Human Development cosponsored a Research Summit, Dosing in Children With Injured Brains and Cerebral Palsy.³⁶ The proceedings from this summit have been published, and Jill C. Heathcock, MPT, PhD (the principal investigator [PI]) was part of the planning committee of the summit and one of the authors on the manuscript.^36,39 In addition, parents at Nationwide Children's Hospital (NCH) identified intensity as an important aspect of how to improve motor skill development in their children with CP, and the parent CP group Reaching for the Stars and Counting Possibilities also identified intensity and service delivery as areas of importance. In these and other ways, dosing of therapy services has been identified as important by parents/caregivers, clinicians, and research groups.⁶⁴

Impairment in motor development is the primary reason parents seek, or a child receives medical referral for, PT services.³³ Patients as well as parents and caregivers of children with CP identify motor function and independent mobility as the primary outcomes they want to improve. The ACHIEVE study focused on research aimed at improving motor function to the best extent possible. Family-centered care is a foundational principle of pediatric therapy allowing families to make informed decisions about current and future therapy for their children.^65-68 The ACHIEVE study used the infrastructure of an existing and successful clinical program, the expanded implementation of a first-ever RCT with a nonrandomized (self-selected) cohort on intensity in an outpatient setting, and innovative data analytic procedures, including goal attainment scaling (described in study outcomes) that ensures outcomes that are important to parents. We evaluated comparisons between HIP and usual weekly therapy with identical total number of skilled therapy hours in 2 groups. Of particular interest were individual child characteristics that may mediate the effects of each intensity, including GMFCS level (a severity scale in which I is highest function and V is lowest function), level of independence at baseline, and age.³⁹ The results of ACHIEVE are likely to immediately improve the efficiency of clinical practice and improve care with sustained changes in practice.

Conceptual Framework

Pediatric rehabilitation has been guided in recent years by an emphasis on providing activity-dependent treatment types that ideally^63,69 (1) improve activity during daily life at home and in the community to allow for sufficient repetitions of acquired and emerging motor skills; and (2) improve developmental plasticity and neural recovery by providing input to damaged or dysfunctional neurons, thereby diminishing the effects of developmental disregard and potentially improving learning by an increase in brain growth factors.^69-72 In collaboration with others, Dr Heathcock proposed a conceptual framework that identifies potential mediators of dosing.³⁹ We designed the path model for hypothesis-driven comparative dosing studies for children with CP using a comprehensive approach to dosing based on knowledge of child development, CP, and rehabilitation studies (Figure 1). The ACHIEVE project tested the hypothesis that intensity of therapy is a key moderator in improvements in motor function. We specifically tested the hypotheses that HIP therapy results in comparatively greater improvements than does usual weekly therapy. In addition, patient age, level of CP severity, family preference, and characteristics of therapy were evaluated using the path model to determine whether they influenced the effects of dose or were indicative of important subgroups. We incorporated this theoretical model of dosing to guide our planning and compare all parameters of dose, including type and length of treatment and characteristics of the trained therapists, should differences arise. This study (and most of the PT field) is situated within the medical model of disability, in which individuals receive therapy to improve body structures and function to enhance their participation in activities. However, it is important to acknowledge that there is also a social model of disability in which society creates opportunities and limitations for participation in a life situation. It was not within the scope of this study to address such societal structures; our stakeholders were able to provide valuable insight into this area.

Figure 1

Path Model for Dosing^,.

The ACHIEVE study compared the effectiveness of 2 intensities of treatment for children with CP in an outpatient PT setting. HIP therapy consisted of 2 hours of focused and high-intensity treatment per weekday for 4 weeks. In comparison, usual weekly therapy spreads out the therapy over time, with a total of 40 hours of treatment spread over 40 weeks. Usual weekly therapy is considered the standard of care, although both dosing models are used clinically for children with CP.^1-3 Importantly, the total number of hours and daily hours of prescribed therapy were the same for the 2 groups in ACHIEVE, allowing for a powerful and scientifically credible study because the compared regimens differed only in the distribution of the hours. Most children with CP receive usual weekly therapy treatment. The ACHIEVE study aimed to provide insights into issues related to rehabilitation, dosing, and high-intensity programs for children with CP, including, “What are my options for intensity of therapy, and what are the benefits and harms of higher intensity and usual weekly programs?” In addition, it was designed to answer the questions, “Given my child with CP's personal characteristics, conditions, and preferences, what changes should I expect in my child if I enroll in a HIP or usual weekly therapy program?” and, “How do I (a parent) estimate my child's skills and how does that contribute to changes in motor function?” Ultimately, the answers to these questions will vastly improve how patients and their caregivers make better-informed health care choices.

Planning the Study

Planning ACHIEVE took several years with many overlapping components. Parents of younger infants with CP and nonambulatory children approached The Ohio State University (OSU)/NCH therapists about the need and desire for a HIP program. They wanted to see whether motor function in their children could be “improved quickly” and whether they could “come to therapy all at once.” They asked, “What is the best intensity of therapy?” and “When should I start intense therapy for my child?” Clinicians reviewed the available high-intensity programs at NCH and found programs for school-age children using a summer camp model or those for higher-functioning children that focused on fine motor skills of the hand or walking on a treadmill; however, they did not find programs for younger children or children who were more severely affected that focused on gross motor function. Researchers and clinicians in Dr Heathcock's laboratory reviewed the literature and found efficacy for HIP models and weekly models of therapy in the younger population. A HIP clinical therapy program (called STRIVE) was designed and implemented at NCH by Rachel Ferrante, DPT (co-investigator [co-I]), for younger children at all severity levels with CP and other pediatric diagnoses. Researchers continued to review the literature; participate in consensus conferences with parent and patient stakeholders; identify gaps in comparative effectiveness of dosing and intensity, especially for young children with CP; and apply for and receive funding to begin to study dosing.

The ACHIEVE team was put together to evaluate HIP programs for children with CP and to seek funding for comparative effectiveness studies. An adult patient with CP was part of our stakeholder panel, and all partners contributed to study planning, designing the intervention, and choosing the outcomes that were important to them. During the development of the original PCORI application, all stakeholders, and especially parents and scientists, were heavily involved in deciding the primary and secondary outcome domains and creating feasible dosing parameters for families with young children with disabilities. The role of the parent stakeholders continued during our resubmissions when we added a play outcome measure that was suggested by a parent stakeholder. Our stakeholders were all involved in discussion and planning and came to group consensus regarding the addition of a nonrandomized cohort and reduction of the overall sample size. One of the primary reasons for families declining participation was their inability to commit to the time and scheduling requirements for study participation. We added a nonrandomized cohort in December 2017 to increase our recruitment numbers and allow families to participate in the study that would have not been able to do so if randomization were required. Patient and stakeholder partners continued to participate in study planning and design by (1) attending a monthly team meeting; (2) participating in quarterly advisory discussions; (3) preparing reports and disseminating information to other families; (4) providing input on any changes made during the study, especially regarding dosing parameters; and (5) planning for dissemination.

Conducting the Study

Patients and stakeholders participated in the study by monitoring the conduct of the project, assisting with recruitment of study participants, drafting the informed consent for readability, helping draft survey question and advisory panel topics, reviewing progress and challenges, advising the whole team regarding stakeholder perspectives, and planning for dissemination of the results to other families.

Disseminating the Study Results

Patients and stakeholders identified partner organizations for dissemination, planned dissemination efforts, and participated in dissemination efforts in oral and written forms during advisory panel meetings and scientific meetings. Our parent team is drafting an article for Columbus Parents Magazine (identified as a possible partner organization for dissemination in a revision by a stakeholder) to disseminate study findings and communicate them in an understandable and usable way. We have also partnered with PT in Motion Magazine for a nontechnical summary of pragmatic and patient-centered pediatric research (magazine article submitted to Editor Donald Tepper who reported it was suitable for publication).

Principles for Engagement

1. Reciprocal relationships: Our governance plan articulates the specific roles and responsibilities for the research team and stakeholders and outlines tasks, roles, responsibilities, deadlines, and upcoming events. This plan identifies the description of work, ownership of data, criteria for authorship, and definitions of who is responsible for which areas of the ACHIEVE study. It establishes rules for appropriate usage of laboratory facilities and interaction with children with CP, caregivers, research staff, and medical professions.
2. Co-learning: Members of the ACHIEVE study all have human participant protection training in research study design. Co-learning occurred during all engagement activities. A discussion among all panel members is a key time for stakeholders to contribute and present information.
3. Partnership: All allocation of resources is based on participation at an equivalent hourly rate.
4. Trust, transparency, honesty: These are core values of ACHIEVE. We have found that regular, honest, and, most importantly, genuine communication and interest are key to making major decisions that are made inclusively, and that information is thus shared readily with all research partners.

Identification of Key Stakeholders

An overwhelming strength of the ACHIEVE project was our patient and stakeholder engagement. Our team of stakeholders, patients, and researchers continues to be motivated and determined to improve the decision-making and care of how to deliver the best intensity of therapy to improve outcomes in children with CP. Stakeholders have been involved in all aspects of this project. To identify key stakeholders for the ACHIEVE study, we followed the 3 steps described below.

Step 2: Identifying Community Partners

Community partners were chosen based on skills and resources; assets that already exist in the community; expertise in patient-centered needs in pediatrics; comparative effectiveness research; and dosing, credibility, and end users of the research. Table 2 lists our community partners and their contributions. OSU served as the lead agency, and Dr Heathcock was responsible for orientation meetings and one-on-one discussion with community partners and stakeholders; she ensured stakeholder involvement in the planning process and compiled shared goals and dissemination of research findings to parents, referring physicians, and therapists.

Table 2

Community Partners.

Description of Engagement Type

Stakeholders were engaged in formulating research questions; defining essential characteristics of children with CP who respond well to each model of intensity, comparators, and outcomes; monitoring ACHIEVE conduct and progress; and disseminating research results. To accomplish this, we implemented 5 types of engagement. A summary of each event category appears in Table 3.

Table 3

Annual Advisory Panel Discussions (in Addition to Monthly Meetings).

1. Kickoff – a meet and greet for all identified community partners and individual stakeholders and advisory panel meeting. Set dates for data safety and monitoring board (DSMB) meetings, IRB training and submission, plan materials for initial recruitment, identify and invite new stakeholders, plan treatment fidelity training sessions.
2. One-on-one meetings with the PI – individual and small-group meetings and commitment from new stakeholders. Assess strengths and interests of individual stakeholders and assign/choose tasks accordingly.
3. Monthly meeting – Whole-team meeting for ACHIEVE study business in person and on conference calls. Parent stakeholders meet with the entire research team (encouraged) and separately a few times per year. Review recruitment, milestones, and reliability of outcome assessors and missing data. Set teams for dissemination through abstracts and poster presentations and drafts of community pamphlets and newsletters.
4. Advisory panel discussion – small- and large-group discussion on ACHIEVE and future patient-centered research projects. Dosing, recruitment and retention, and dissemination are likely topics of discussion.
5. Advocacy walks and volunteer activities – an advocacy event for CP (sponsored by Counting Possibilities, March of Dimes, and the Dublin Special Education Advancement Council), followed by an advisory panel meeting. Review yearly progress and DSMB minutes, solicit feedback, make changes as needed, and review and update milestones for following year. The March of Dimes walk and a volunteer activity at the Ronald McDonald House were added to this category based on stakeholder feedback.

Study Overview

The overall goal of this project was to compare the effectiveness of HIP and weekly therapy in treating children aged 2 to 8 years with CP for motor rehabilitation. The outcomes were gross motor function (primary outcome: Gross Motor Function Measure-66 [GMFM-66] score), patient and parent engagement, development in 5 domains, spontaneous play, health-related QOL, and family satisfaction with care. The study was informed by motor learning theory, which allows for a framework to ensure and measure treatment fidelity, and a theoretical path model for dosing in children with CP. It also leveraged existing clinical practices at a major children's hospital (NCH) and its satellite sites. The aims were as follows:

1. Conduct an RCT with a nonrandomized option. Participants were children aged 2 to 8 years with CP. There were 46 children in the RCT and 54 children in a nonrandomized (family choice) group that compared the short-term effects and explored the long-term effects of 2 service delivery models of outpatient PT: usual weekly and HIP. Usual weekly was defined as therapy for 1 hour per day, 1× per week for 40 weeks. HIP therapy was defined as 2 hours every weekday for 2 periods of 10 consecutive weekdays (total of 4 weeks) of a repeated periodic bout, with an 18-week period between bouts. The 9-month assessment was the primary end point of the study.
2. Provide insights into the issues related to providing rehabilitation services for children with CP by determining individual differences in children's responses to treatment and differences in parent preferences.
3. Monitor progress and variability of motor function.
4. Explore patient and family factors associated with treatment response.
5. Identify relationships between parent assessment of function and gross motor function level as determined using standardized assessments delivered by a therapist.

We hypothesized that children would make greater gains participating in HIP therapy than with usual weekly therapy. Children who are younger and more engaged, have lower GMFCS levels, and have higher levels of parent engagement and satisfaction were expected to make greater gains. Frequent measures of motor function were expected to have within-child and between-child variability given what is known about the heterogeneity of (1) child development and (2) motor function of children with CP across GMFCS levels.

Study Setting

The study was performed at 7 satellite sites affiliated with 1 institution in central Ohio, comparing 2 intensity protocols for children with CP, while keeping the total hours of therapy the same. We accomplished this by building on an active, productive clinician-researcher-stakeholder collaboration on treatment dosing in children with CP, with extensive clinical and scientific expertise in PT, motor development, neurology, and pediatric motor rehabilitation. We chose NCH, which is one of the largest pediatric hospitals in the country, because the infrastructure for enrolling, wait listing, and treating children aged 2 to 8 years with CP using a HIP and usual weekly treatment protocol for the proposed PCORI study was already in place. We have been enrolling and treating children with CP and other serious motor disabilities in our clinical high-dosing program, called STRIVE, since 2012.^75,76 For the ACHIEVE study, most patients were seen at the main hospital location due to its centralized location in the city, the size of the treatment space, and the number of therapists involved in treatment of the study patients. This location also offered access to temporary housing through the Ronald McDonald House of Central Ohio for those families needing to travel from their homes to stay closer to the location during their child's treatment period. The main hospital PT department is in the outpatient section of the hospital located in an urban setting of Columbus, Ohio. The patients treated at this location span a variety of racial and socioeconomic backgrounds.

All other locations within the NCH rehabilitation system were also used as treatment locations based on convenience and preference of the patient families. These “close to home” sites are located closer to suburban neighborhoods around the greater Columbus area and include locations in Dublin, Lewis Center, Westerville, East Columbus, Hilliard, and Clintonville, Ohio. These locations serve a variety of patients with diverse racial and socioeconomic backgrounds. One additional “close to home” center in Ontario, Ohio, approximately 70 miles north of NCH's main location serves a more rural area; however, it is still a diverse socioeconomic population. Use of the various locations spanning Franklin County and extending into the northern counties of Ohio, along with the access to temporary housing through the Ronald McDonald House, allowed patients from a wide variety of locations around Ohio and bordering states to participate in the study.

Participants

Target Population

The target population for the study was children with CP at all GMFCS levels from 2 to 8 years of age, who were randomly or nonrandomly (self-selected or parent choice) assigned to 1 of 2 groups: HIP or weekly therapy. The overall study design can be found in Figure 2. An average of 4 patients every month were recruited, with an overall goal of 128 children: 120 verbally agreed to participate, with 100 starting treatment. There is comprehensive diversity of ethnicity, race, and socioeconomic status (SES) of patients among the sites from which these patients were recruited.

Figure 2

Study Design for ACHIEVE.

Interventions and Comparators or Controls

The treatment phase of this study design was 40 weeks (9 months), for a total of 40 hours of one-on-one therapy for the 2 groups. The total time of treatment was carefully chosen, taking into account several factors. First, previous work using high-intensity therapy to develop upper-extremity skills for adults and children with stroke demonstrated efficacy from 30 to 90 total hours, with 90 hours showing more benefits. We used the lower end of the total treatment hours in these studies as a guide for the minimum and range of total treatment hours proposed in this project because of the differences in attained ability between upper-extremity training, which uses only small muscle groups, compared with fine motor therapy and motor skill training, which uses primarily large muscle groups and the whole body. Second, our high-intensity clinical program (STRIVE) and published literature on HIP treatment using pre and post outcomes at ∼40 hours of therapy show efficacy for improvements in motor and cognitive skills for children with CP who are aged 2 to 8 years. Third, it is typical for children in this age range to receive continuous therapy for most of the year. Fourth, based on our previous clinical research experience running usual weekly and HIP programs, we anticipated an 85% attendance rate because of the potential for illness and holidays to overlap therapy appointments. The treatment period for both groups was 9 months (40 weeks). We anticipated the average amount of one-on-one therapy to be 34 hours for each group. For the above-listed reasons, we chose 40 hours as our prescribed dose (allowing for 85% completion of 40 total hours to be 34 hours) of therapy for patients in the proposed project. The total amount of therapy was equivalent between groups, and we compared the following 2 dosing schedules: usual weekly (1 hour per day, 1× per week, for 40 weeks) and HIP (2 hours every weekday for 2 bouts of 10 consecutive weekdays, for a total 4 weeks; bouts were separated by an 18-week period).

All therapists who were involved in treatment throughout the ACHIEVE study were licensed physical therapists with experience in the treatment of pediatric patients specifically with a diagnosis of CP. All therapists completed specific training in (1) principles of motor learning, (2) providing feedback and shaping behaviors, (3) adherence to dose, (4) the importance of repetition of movements, and (5) the ACHIEVE protocol. Although examples were used to illustrate these points, specific treatment activities (such as exercises or games) were not prescribed. Instead, therapists selected and adapted activities during treatment sessions to address each participant's goals. Therapists provided cueing and feedback to participants throughout each treatment session using the principles of motor learning that were described in fidelity training. Heart rates were monitored to measure intensity and fatigue, and the videotaped sessions were used to measure patient engagement via the observations described previously. Therapists modified activities in the moment as needed based on this information. Only therapists with demonstrated positive treatment fidelity checks provided treatment to patients enrolled in ACHIEVE. All therapists met minimum fidelity standards over the treatment period by using a goal-directed approach, using active functional movements, adhering to principles of motor learning, providing shaping and feedback, adhering to dose, and focusing on repetition.

Study Outcomes

Outcome evaluations (OEs) were performed by blinded evaluators at baseline (month 0, before treatment), at month 9 (end of treatment) for patients to assess short-term effects, and at months 12 and 18 for some patients (with 18-month visit time before study end) to assess long-term effects. OE and PT consultation (PTC) sessions occurred during the same visit when they overlapped. Parent surveys were collected at 0, 4.5, and 9 months.

The GMFM-66 was our primary outcome measure for motor function, and motor function in general was identified as most important by parents. The GMFM has 2 versions. The GMFM-88 has 88 items, and the GMFM-66 has 66 items that can be calculated entirely from the GMFM-88. Both instruments have scores ranging from 0 to 100, and higher scores indicate better motor function. Our preliminary data and estimates of power were based on the GMFM-66. There is some emerging evidence that the GMFM-66 may underestimate change in younger children and in children classified at higher GMFCS levels (ie, lower functioning), because most of the eliminated items come from the section on lying and rolling. Thus, we tested all 88 items on the GMFM in our outcomes protocol because our age range was 2 to 8 years, and we include all severity levels; the 88-item version includes more items that are appropriate for younger children and those with more severe motor impairment. Besides the GMFM scores, patient response was classified as “not improved,” “improved,” and “greatly improved” for those with a post-to-pre change in GMFM-66 score of <1.58, 1.58 to 3.71, and >3.71, respectively, as per the literature.⁷⁸

Goal attainment scaling, a novel type of OE used in ACHIEVE, needs further explanation. As a framework, goal attainment scaling asks the patient and family to set the criteria for improvement by establishing activity or participation goals that reflect what the patient and family consider meaningful (ie, goal attainment scaling is inherently patient centered). Furthermore, goal attainment scaling is an established clinical method for quantifying the achievement of goals and offers a forward-thinking approach for measuring meaningful change in patient-specific activity or participation goals for pediatric patients. For the reasons highlighted above, we used goal attainment scaling as a secondary end point and a way to measure the effectiveness of treatment intensity in all groups in the ACHIEVE study and to compare between and among treatment groups. A change score of 2 is expected on the Goal Attainment Scale (GAS).

The rest of the secondary outcomes are summarized in Table 4.

Table 4

Outcome Evaluations.

Sample Size Calculations and Power

We planned to enroll approximately N = 108 participants, including both the randomized cohort and the nonrandomized cohort. Our primary hypothesis was that HIP treatment could provide better short-term improvement (baseline to 9 months) in GMFM-66 scores than could weekly treatment. From our preliminary data, the HIP treatment group had a standard effect size of >1.35 improvements, and the weekly treatment group had a standard effect size of <1 (we used an effect size of 0.75 for the weekly group in our revised protocol). Originally, we proposed to enroll 130 participants per group for an RCT to provide at least 80% power to detect an effect size difference of 0.35. In the final approved protocol, a sample size of 45 participants per group provides at least 80% power to detect an effect size difference of 0.60 at a significance level of .025 (1-sided test based on a 2-sample t test). The power to detect various effect sizes between 2 treatment groups with different sample sizes at a significance level of .05 is illustrated in Figure 3. We planned to recruit 54 participants for each group (for a total of 108) to account for attrition of up to 9 participants per group. To reduce the inherent variability and achieve comparable numbers of participants between the HIP and the usual weekly treatment groups, for those families willing to be treated in the randomization cohort, we randomly assigned participants at a ratio of 1:1 into 2 groups using a stratified procedure based on age (2-3, 4-5, and 6-8 years), sex (male/female), and GMFCS level (levels I-II, level III, or levels IV-V) at baseline. We also included a cohort of patients with a self-selected treatment group (weekly vs HIP) based on their family circumstance (time commitment and scheduling issues) with pooling of groups.

Figure 3

Power of Detecting Different Effect Sizes With Sample Sizes of 45, 60, 80, and 100 Participants Per Treatment Group.

Time Frame for the Study

The intervention period for both conditions was 40 weeks (9 months), although the intensity of therapies differed as described previously. The length of time is important to note because it is equivalent in both groups, allowing for the first-ever comparison of effectiveness of these 2 interventions. In addition, all patients enrolled in ACHIEVE had 1 hour per month of PTCs during the treatment period. A subset of the PTCs were videotaped. PTCs had a heavy focus on home exercise programs (HEPs) and were provided by research personnel who were blinded to group assignment. Monthly PTCs were included for 3 reasons. First, time dedicated to HEPs and parent education is a nonstandardized component of outpatient PT. We aimed to provide as much standard one-on-one treatment time with patients and therapists as possible so that we could define our dose as 40 hours of one-on-one therapy. Taking the nonstandardized home exercise component out of skilled therapy time helped accomplish this. The research physical therapist who provided the PTC had access to the initial evaluation and goals set by the treating therapist, all data from the outcome measures (OEs), and 1 hour with the patient and family each month, allowing for a comprehensive PTC and HEP. Second, given the nature of HIP therapy, patients had several weeks “off” where they were not receiving treatment. Although unlikely, these off weeks could have potentially led to an unanticipated decline in motor skills. Motor skills were monitored during these monthly sessions. Importantly, the HIP group participants had more therapy before the off weeks than did those in the usual care group. After receiving feedback from our parent and clinician stakeholders, we decided that this break could be too long and that providing monthly PTC was an acceptable solution because it provided regular contact with a physical therapist to progress the HEP and monitor skills. As such, participants assigned to the HIP group did not go more than 4 weeks without motor skill monitoring, consultation, and general home therapy program progression. Third, monthly assessment and checking in with families during the treatment period is a strategy for improved retention among all ages and GMFCS levels and ensured that study data were collected thoroughly and systematically from all study participants; it was also a suggestion by our parent stakeholders. Fourth, monthly visits allowed for tracking of the natural history of motor skill development, including skill progression and variability during the treatment period, key features that may impact rehabilitation in children with CP.

The follow-up schedule also included outcome assessments performed by blinded evaluators at baseline (month 0, before treatment), at month 9 for participants to assess short-term effects, and at months 12 and 18 for a subset of participants to assess long-term effects. OE and PTC sessions occurred during the same visit. Parent surveys were collected at 0, 4.5, and 9 months only. Blinded evaluators were physical therapists, students, research assistants, and postdoctoral fellows in Dr Heathcock's laboratory who had no contact with ACHIEVE participants when they were at NCH for therapy. PTCs and OEs occurred in Dr Heathcock's pediatric laboratory at OSU or in the participant's home. We anticipated that a third of the families enrolled in ACHIEVE would choose to come to the laboratory and that two-thirds of the families would choose to have us come to their home. Fifty-nine percent of the assessments were completed in the laboratory, and 41% were completed in the home. OEs were conducted in a consistent location for an individual child, using the same transportable testing equipment regardless of location, ensuring consistency of testing.

Data Collection and Sources

The baseline assessments were scheduled by the study coordinator at the convenience of the family. Follow-up appointments were scheduled by the research therapist, if possible, immediately after an assessment. If necessary, research therapists would call families to schedule or would contact them via text message. A text message was also sent as an appointment reminder the day before a scheduled research visit. Secure, encrypted email was also used if an email address was provided by the family. If families did not respond to any of the 3 scheduling attempts, the study coordinator mailed a letter to the family to inform them that we were unable to reach them for scheduling and requesting that they contact us if they wished to remain in the study. If no response was received by the family, we assumed they had no further interest in participating in the study and marked the study participant as “withdrawn.”

When a family contacted us with the wish to withdraw, we asked if they were comfortable sharing the reason for their decision. If possible, we offered solutions to alleviate the issue and keep the family enrolled (eg, offer transportation or different treatment times) or offered to reenroll the family at a later date if continued participation in the study was not possible at that time (eg, because of surgery).

We improved retention through several methods. Families were given the option to complete questionnaires online (via REDCap), minimizing the additional time they needed to spend at the assessment visits. Via newsletters, we shared study information and resources for families raising children with CP. Families received $20 per OE visit and PTC visit; the child also received a toy at each of those visits. Participating children and stakeholders received an ACHIEVE study T-shirt.

Analytical and Statistical Approaches

The primary hypothesis of this study was that HIP treatment would lead to better short-term improvement in GMFM-66 scores (baseline to 9 months) in the HIP group than would usual care in the weekly treatment group. Long-term effects and parent choice effects (randomized vs nonrandomized) were also explored. Secondary and exploratory outcomes included scores on the GMFM-88, the GAS, and the Bayley Scales of Infant and Toddler Development, Third Edition (Bayley-III), as well as on several parent-report surveys (Pediatric Outcomes Data Collection Instrument [PODCI], Pediatric Evaluation of Disability Index [PEDI], and Family Professional Partnership Scale [FPPS]).

We created a DSMB at OSU in collaboration with the Center for Clinical and Translational Sciences, which provides services for data and safety monitoring (DSM). The board included Dr Noritz (physician, NCH); Jeff Pan, PhD (biostatistician, OSU); Dr Amy Bailes (physical therapist, Cincinnati Children's Hospital); and Drs Sharon Ramey and Stephanie DeLuca (both in developmental psychology, Fralin Biomedical Research Institute, Virginia Tech). None of the members of the DSMB were associated with the study. All members of the DSMB have experience in child development and CP. This board convened at the start of the study and yearly after that. Three additional meetings also occurred. The first meetings occurred after 8 participants completed therapy, and the data were reviewed. The second meetings occurred when the nonrandomized cohort was considered. The nonrandomized cohort was considered for several reasons, including our experience with a lower-than-expected recruitment rate, ongoing recommendations from the ACHIEVE stakeholder advisory board, and review of the surgical literature in which patients and providers cannot be blinded to group assignment. The third meeting occurred midway through the trial to see if there was a subgroup of children who were not responding. The DSMB provided a list of types of data they would like to review and at what intervals they would like data reports on immediately posttherapy, short-term, and long-term outcomes.

Because of the addition of the nonrandomized cohort, we revised our data analysis plan, which only consisted of the 2 treatment groups from the RCT. The revised data analysis plan in the report includes all 4 groups: HIP or weekly treatment in the RCT or in the nonrandomized cohort. An evaluation of the cohort effect, randomization or self-selected, was added to the analysis plan before data lockdown. The primary comparison of the treatment effect was based on all 4 groups using a linear mixed model, as detailed below.

All baseline data were summarized using descriptive statistics, with means ± SD reported for continuous variables, and frequencies and percentages reported for categorical variables. Summaries were generated for all patients by treatment group for the randomized and nonrandomized cohorts separately. Most outcome measures were treated as continuous measures in this study. For each outcome measure, individual plots over time were generated for each group to visually inspect the trend of change and potential data outliers before data analysis. All outcomes were summarized at each time point by cohort and treatment.

Our primary analysis was conducted using the intention-to-treat (ITT) principle and included all consented participants with valid baseline measurements. SAS v9.4 (SAS, Inc) was used for our analysis. Although we have made every effort to minimize the missing data for this study, missing data are expected for various reasons. During each OE or PTC visit, the blinded evaluator visually inspected the fields on all outcome measures to ensure completeness. The pattern of the missing data was reported biannually during the study within REDCap and at the end of study. Our primary analysis used mixed models which, under the assumption of missingness at random (MAR), use maximum likelihood estimation to deal with any missing data. Sensitivity analyses were also conducted based on the completed data set (per protocol) and data sets in which multiple imputation was used to impute missing outcomes.

Linear mixed models for repeated measures were used to compare the effectiveness of the 2 treatment groups in the change from baseline to 9 months (primary, short-term effect) and 12 and 18 months (secondary, long-term effect) after adjusting for the effects of stratification factors as potential covariates. Given that only 37 participants had an 18-month follow-up—less than 50% of the ITT population—the primary analysis only looked at the baseline, 9-month, and 12-month data. An unstructured variance covariance matrix was assumed for the repeated measures within each participant when appropriate, and we used alternative models that included autoregressive (AR)(1) and compound symmetry models. The original fixed effects included in the model were time, treatment, and their interaction (time × treatment). Cohort (randomized/self-selected) and the interaction of cohort with treatment (cohort × time × treatment) was also considered as a fixed effect in the primary analysis. The estimated changes of the outcome over time were plotted for 4 groups, which were based on the combination of cohort (randomized/self-selected) and treatment (HIP/weekly). The 9-month (40 weeks) short-term treatment effects were reported for each of these 4 groups based on the corresponding contrast of the time for each group from the model along with their 95% CI. If the interaction of cohort, treatment, and time (cohort × time × treatment) was not statistically significant, the primary hypothesis test of the short-term treatment difference between HIP and weekly treatment was estimated based on the corresponding contrast of the parameters of time × treatment in the model, which were estimated based on all participants from the RCT cohort and the nonrandomized cohort. Otherwise, it will be based on the corresponding contrast of the parameters of the cohort × time × treatment in the model. In addition, if the time × treatment interaction was also not significant, an overall short-term treatment effect was estimated based on the contrast (ie, a linear combination of variables [parameters or statistics] whose coefficients add up to 0, allowing a comparison of different treatments) of the time parameter (from baseline to 9 months) from the model, which was the “overall treatment effect” estimated using data from all 4 groups.

All primary and secondary outcomes that can be treated as continuous variables were analyzed using the same approach described above. Outcome measures include the GMFM, Bayley-III, GAS, PODCI, and PEDI-Computer Adaptive Test (PEDI-CAT).

For aim 1, to control for type I error, we first tested the hypothesis that HIP treatment was superior to weekly treatment in its effect on short-term GMFM score at a 1-sided significance level of α = .025 (or 2-sided test at α = .05). Because the HIP treatment was not significantly different from the weekly treatment, we tested the difference of the 2 treatment groups for the long-term effect and other secondary outcomes as exploratory analyses. For aim 2, we investigated the association of the short- and long-term effects vs patients' baseline characteristics (such as baseline GMFCS level, age, sex, race), parent engagement, patient engagement, and parent/family characteristics (income level).

In addition to the statistical analysis conducted above, we plan to develop a multivariate linear regression analysis (GLM SELECT procedure in SAS) to predict the short- and long-term effects based on these explanatory variables in the near future. Interaction of the treatment with disease severity and sex could be included in the model selection procedure. In addition, exploratory analysis will be conducted to illustrate the relationship of different factors using a structural equation model based on the path model (as in Figure 1).

Aim 1

The primary hypothesis of this study was that HIP treatment would provide better short-term improvement (baseline-9 months) than would usual weekly treatment in terms of GMFM-66 scores in children with CP (Figure 5, Table 7).

Figure 5

Primary Outcome: GMFM-66 Scores From Baseline to 12-Month Follow-up (Scale, 0-100).

Table 7

Short-term Effects From Baseline to 9 Months (End of Treatment) on GMFM-66 Score for Each Group and Treatment Effect Differences Between HIP and Weekly Therapy.

At the time of this report, of the 100 eligible patients assigned to treatment (randomly assigned, n = 46; self-selected, n = 54), 76 (76%) had a 9-month end-of-treatment follow-up assessment (78% and 74% for the 2 cohorts, respectively, with no significant difference between the 2 cohorts; P = .65); 67 (67%) had the 12-month follow-up assessment (74% and 61% for the 2 cohorts, respectively; P = .21); and 47 (47%) had the 18-month follow-up assessment. As such, we excluded the 18-month follow-up (data collection ongoing), and the primary analysis only used the baseline, end-of-treatment (9 months), and 12-month follow-up data. The GMFM-66 score ranges from 0 to 100, where higher scores indicate better motor function. The fixed effect of the linear mixed model for repeated measure included treatment (HIP vs weekly), cohort (randomized vs self-selected), time (baseline, 9-month follow-up, and 12-month follow-up, as categorical data), the time × treatment interaction, and the time × treatment × cohort interaction. The AR(1) model was selected to account for the association of the measures from the same individual at different time points. The interactions of time × treatment × cohort and time × treatment were not statistically significant (P = .28 and .56, respectively), which suggested that differences in the change over time among different treatment groups and different cohorts were not significantly different.

Figure 5 illustrates the estimated GMFM-66 score at different time points based on the linear mixed model. All 4 groups had significant improvement at the end of the 9-month treatment period (all P ≤ .005, Table 7), and the overall improvement in GMFM-66 score from baseline to 9 months was 4.19 (95% CI, 3.23-5.14; P < .0001). The difference between the HIP and weekly treatments was not statistically significant (−0.92; 95% CI, −2.83 to 10.98; P = .34), whereas the weekly treatment group had statistically nonsignificant greater improvement than did the HIP group. The direction of this nonsignificant difference in improvement is different from our hypothesis, because we expected the difference between HIP and weekly to be positive by 1 to 3 points. Sensitivity analyses were conducted using different covariance structural models and including other covariates, such as age; these yielded the same conclusion that the treatment effects (difference of the 9-month short-term effect between HIP and weekly treatment) were all not significant, while all 4 groups had significant short-term improvement from baseline to the end of treatment at 9 months.

The same model and sensitivity analysis approaches were used for all other outcomes summarized in this report.

Secondary Outcomes

With the secondary outcome of GMFM-88 score, we used the same model as for the primary outcome. Although all 4 groups had statistically significant improvement at the end of the 9-month treatment period (Figure 6 and Table 8), the difference between the HIP and weekly treatments is not statistically significant (−0.70; 95% CI, −3.02 to 1.63; P = .55), where participants in the weekly therapy group had nonsignificant greater improvement than did those in the HIP therapy group. The direction of this is different from our hypothesis.

Figure 6

Secondary Outcome: GMFM-88 Score (Scale, 0-100).

Table 8

Secondary Outcome: GMFM-88 Score.

Parent-reported measure of mobility (PEDI-CAT)

While overall the 4 groups had a statistically significant improvement at the end of the 9-month treatment period (1.35; 95% CI, 0.58-2.13; P = .0008; Figure 7), for each individual group, only the 2 self-selected groups had statistically significant improvements (Table 9). The difference between the HIP and weekly treatments was not statistically significant (0.16; 95% CI, −1.39 to 1.71; P = .84).

Figure 7

Parent-Reported Measure of Mobility: PEDI-CAT Score.

Table 9

Parent-Reported Measure of Mobility: PEDI-CAT Score.

PEDI-CAT measure of social and cognitive domains

Overall, the 4 groups combined had significant improvement at the end of the 9-month treatment period (1.09; 95% CI, 0.48-1.71; P = .0006); however, not all 4 groups individually had statistically significant improvement at this time (Figure 8 and Table 10). The difference between the HIP and usual weekly treatments was also not statistically significant (0.25; 95% CI, −0.98 to 1.47; P = .69).

Figure 8

Parent-Reported Measure of Social and Cognitive Domains: PEDI-CAT Score.

Table 10

Parent-Reported Measure of Social and Cognitive Domains: PEDI-CAT Score.

PODCI global function

Overall, the 4 groups had no significant improvement at the end of the 9-month treatment period (0.73; 95% CI, −1.23 to 2.68; P = .46), and none of the individual groups had statistically significant improvement at this time (Figure 9 and Table 11). The difference between the HIP and usual weekly treatments was not statistically significant (−0.17; 95% CI, −4.07 to 3.73; P = .93).

Figure 9

Parent-Reported Measure of Global Function: PODCI Score.

Table 11

Parent-Reported Measure of Global Function: PODCI Score.

PODCI upper extremity and physical function

Overall, the 4 groups had significant improvements at the end of the 9-month treatment period (Figure 10; 3.31; 95% CI, 0.90-5.71; P = .0075); however, none of the individual groups had statistically significant improvement at this time (Table 12). The difference between the HIP and usual weekly treatments was not statistically significant (0.44; 95% CI, −4.38 to 5.25; P = .86).

Figure 10

Parent-Reported Upper-Extremity and Physical Function: PODCI Score.

Table 12

Parent-Reported Upper-Extremity and Physical Function: PODCI Score.

Bayley-III gross motor skills

Overall, the 4 groups had significant improvements at the end of the 9-month treatment period (3.84; 95% CI, 1.83-5.85; P = .0004; Figure 11); however, not all of the 4 groups had statistically significant improvement at this time when we analyzed each group individually (Table 13). The difference between the HIP and weekly treatments was not statistically significant (−1.11; 95% CI, −5.12 to 2.91; P = .58).

Figure 11

Secondary Outcome: Gross Motor Skills, Bayley-III Score (Scale, 0-72).

Table 13

Secondary Outcome: Gross Motor Skills, Bayley-III Score.

Bayley-III fine motor skills

Overall, the 4 groups had significant improvements at the end of the 9-month treatment period (2.21; 95% CI, 0.59-3.84; P = .009; Figure 12); however, none of the 4 groups had statistically significant improvement at this time when we analyzed each group individually (Table 14). The difference between the HIP and weekly treatments was not statistically significant (−1.08; 95% CI, −4.33 to 2.16; P = .50).

Figure 12

Secondary Outcome: Fine Motor Skills, Bayley-III Score (Scale, 0-66).

Table 14

Secondary Outcome: Fine Motor Skills, Bayley-III Score.

Family Survey of Satisfaction

At the end of 9 months, FPPS surveys (with 18 questions) were collected from 79 participants. The FPPS scores range from 1 to 5, with 5 being the highest satisfactory level. There was no statistical difference in the average FPPS ratings among the 4 groups (Table 15; P = .84), based on an analysis of variance model.

Table 15

Family Professional Partnership Scale.

GAS Score Obtained at Each PTC and at 9, 12, and 18 Months

There were significant improvements for all 4 groups at the 9-month visit with both the mean GAS activity (Figure 13) and participation (Figure 14) scores. However, the changes in GAS scores over time were not significantly different among the 4 groups. For Figures 13 and 14, change score is on the y-axis, and a score of 2 is expected by the end of the treatment period (9 months).

Figure 13

Secondary Outcome: Change in GAS Activity Score.

Figure 14

Secondary Outcome: Change in GAS Participation Score.

Aim 2

We aimed to provide insights into the issues related to providing rehabilitation services for children with CP by determining individual differences in children's responses to treatment and parent preferences. The characteristics of the participants and their families are described in Table 5. The results indicate that those with lower GMFCS levels (ie, better function) and those with higher parent satisfaction made greater gains in motor function. SES, measured with the Hollingshead Index, and age were not related to greater gains in motor function.

Aim 3

We aimed to monitor progress and variability of motor function during PTCs. During the 8 PTC sessions, data on variables from the GMFM and GAS were collected (Table 16). The variability of these measures was evaluated using the coefficient of variance (COV), which is the within-subject's SD divided by the overall mean values. We also estimated progress using models that were unadjusted and adjusted for the changes of time, and both models yielded similar results, with slightly decreased COV (data not shown). They all showed significant improvement over time (test of linear trend line).

Table 16

Variability of Motor Function.

Aim 4

In this aim, we explored patient and family factors associated with treatment response. GMFM changes from baseline to 9 months are strongly associated with baseline GMFCS levels, GAS score (end of study), and FPPS (Table 17). Participants with lower GMFCS levels had greater GMFM changes (improvement). Also, these participants with greater improvement reported higher goal attainment scaling and satisfaction.

Table 17

Participation and Partnership: Association of the Change in GMFM-66 Score From Baseline to End of Treatment (Variable 1) vs Demographics, Parent Participation, Parent Satisfaction, and Baseline Symptoms (Variable 2).

Aim 5

We aimed to identify relationships between parent assessment of function and gross motor function levels in children with CP. There are strong associations between all measures at baseline and the end of treatment, but no associations between change scores were found for any single instrument (Table 18, Figures 15-17; PODCI_upperex and PODCI_globalfun are from the same instrument). The Pearson correlation coefficients of these variables (variable 1 and variable 2) and their 95% CI and P values are summarized in Table 18.

Table 18

Association of Clinical Outcomes (Variable 1 vs Variable 2) at Baseline, End of Treatment, or Changes From Baseline to End of Treatment.

Figure 15

Baseline GMFM-66 Score and Parent-Reported Outcome.

Figure 16

Posttreatment GMFM-66 Score and Parent-Reported Outcome.

Figure 17

Change in GMFM Score and Change in Parent-Reported Outcome.

Motor Function

The primary patient-centered outcome measure was the GMFM-66, which measures motor function. All groups showed an improvement in GMFM-66 score. In Oeffinger et al,⁹¹ in children with CP who did not experience a change in GMFCS level or surgery over a 12-month period, an average (SD) change of 0.5 (4) points on the GMFM-66 was observed. This value represents the possible “natural history” of expected motor improvement over a 12-month period for children with CP. In the ACHIEVE study, most children improved much more over a shorter time frame in both study groups than did the children in the Oeffinger study. Before the start of this comparative effectiveness study, there was sufficient evidence in the literature, clinical practice, and our own pilot data to suggest that goal-directed pediatric PT using principles of motor learning had established efficacy for improvements in motor function using both HIP and weekly schedules.^{33,35,40,63,92-95} The main finding of our study was that the differences in improvement in motor function between the treatment groups were not statistically significant, although each treatment group had significant improvement at the end of the treatment. Statistically, there was a main effect for time for all groups, no main effect for treatment group, and no group × time interaction. Most patients in all groups reached the MCID, suggesting meaningful clinical improvements. This pattern of results (main effect for time, no main effect for treatment group, and no interaction) was consistent among most of the outcome measures, strengthening our confidence in the primary outcome result.

Dosing of rehabilitation for children with CP remains an important topic for all categories of stakeholders. The literature on dosing, especially regarding high-intensity models, continues to advance.^{15,57,75,96-107} Some of the results from this study were unexpected. Based on our pilot data and the literature, we hypothesized that the HIP group would outperform the weekly group. The major strength of the present study is that we tested 2 different dose schedules but kept the cumulative dose the same, and found that there was no difference between the 2 schedules. However, we acknowledged and continue to advocate for equivalent comparisons in terms of length of time and total number of hours of therapy. The results of this study suggest that 40 hours of outpatient therapy is effective in improving motor function for children with CP,^{17,33,40,108-114} and no significant differences were detected between HIP and usual weekly regimens when the total hours and treatment type were the same. This last point is very important for clinical decision-making for parents, patients, and health care providers, as it means that both frequencies likely work for improving motor function. Scheduling of patients for pediatric rehabilitation is often multifaceted and includes parent and patient preferences, room availability, therapists' availabilities, age of the child, and other patient and family factors. From a pragmatic standpoint, the results of this study suggest that HIP and usual weekly clinical programs are equally effective. There is some literature on CIMT that suggests that if the dose is equivalent, CIMT is not superior to other forms of rehabilitation. Type of therapy (eg, active, goal directed, error based, based on principles of motor learning) and total number of therapy hours, likely at a minimum of 30 hours, are key drivers in producing improvements in motor function in children with CP.^{73,110,115-119}

Although it should be noted that confidence in this conclusion about self-selection is limited by low statistical power (ie, wide confidence intervals with the difference in motor function), this result suggests that families can select which program is most compatible with their schedule, child's temperament, and current goals and priorities for therapy.

The results from the GMFM-88, a secondary outcome measure, mimic those from the primary outcome measure (GMFM-66) demonstrating a main effect for time, no main effect for group, and no group × time interaction. In this study, both the GMFM-66 and GMFM-88 were responsive to treatment effects.^74,79,97,98

Gross motor and fine motor skills were evaluated using the Bayley-III^80,81 as a secondary outcome measure for a subset of children at the appropriate age and skill level. Main effects of time were observed for combined groups, suggesting that there were changes in gross motor and fine motor skills with 40 hours of treatment. However, subgroup analysis demonstrated mixed results for fine and gross motor skill changes, likely because of the smaller sample sizes in some groups. Clinically, OT often focuses more on fine motor skills, whereas PT focuses more on gross motor skills.^17,40,44 In this study, principles of motor learning, goal setting, and functional activities were the key features of the intervention.^63,120 The results suggest that gross motor and fine motor skills improve with this type of treatment and that there is no benefit of one intensity over the other when the total hours are equivalent. Motor function is the outcome discussed for this report. Future work could consider the effects of changes in different domains of development when the FITT principle is used for outpatient pediatric PT. ⁵¹

Parent-Reported Assessment of Motor Function

Parents reported their child's motor function before and after the treatment period using the PEDI-CAT and PODCI.^121-125 The PEDI-CAT mobility subscale is related to the child's gross motor function according to the GMFM and GMFCS level.¹²⁶ In Fragala-Pinkham et al,¹²¹ children with various diagnoses who participated in inpatient rehabilitation achieved mean (SD) gains of 7.9 (6.7) points on the PEDI-CAT mobility domain as observed during their stay (mean, 62.4 days; SD, 70.7 days; mode, 15 days). The PEDI-CAT has good test-retest reliability for children with CP, and scores are not expected to change when it is administered an average of 25 days apart outside the context of treatment. Parents reported a positive change in mobility after the treatment period, and there was a main effect for time. For this measure, parents in the self-selected cohort reported more change than in the randomized cohort, suggesting that parent and family preferences are important considerations for treatment effectiveness and likely for improvements in participation.^17,109,127 The PODCI upper-extremity and physical function subscales were used to measure parent-reported changes in gross and fine motor skills combined. There was a main effect for time with all groups combined but no significant difference for any individual group. This likely means that parents observed changes in motor skills after 40 hours of intervention, suggesting that type and time of therapy are more important to parents than is frequency. For subscales of the PODCI, mean changes of 2.2 to 2.5 points have been observed across a 12-month period for children who maintain their GMFCS level and do not have surgery. Similar to gross motor and fine motor skills measured with the Bayley-III, these outcomes are not as straightforward as the primary and secondary outcomes analysis of gross motor function.

Goal Attainment Scaling

Therapists established activity or participation¹²⁸ goals with parents as an external criterion for improvement that reflects what an individual and family consider meaningful or relevant. Goal attainment scaling in our study was inherently patient centered because the parents helped create the goals.^82,129 As an established clinical method for quantifying the achievement of goals, this monthly assessment during the treatment period offered a forward-thinking approach for the calculation of meaningful change in patient-specific activity or participation goals for pediatric patients. For both activity and participation goals, there was a main effect for time, no main effect for group, and no time × group interaction. This result is consistent with the primary outcome measure, increasing our confidence in the conclusion that the type and amount of therapy are key factors to improvements in function, activity, and participation.^{17,82,108,127,129}

Combining Parent Report and Motor Function

Large, statistically significant correlations were found between parent-reported outcomes (PEDI-CAT and PODCI) and the GMFM and GMFCS scores (Table 18). These results support a strong relationship between parent-reported measures and motor function as measured by clinical professionals. Interestingly, there was no relationship between change score on the parent measure and change score on the GMFM, perhaps due to small sample sizes, yet each demonstrated some treatment effects. Future research could consider both objective and parent reports for similar constructs.

Subpopulation Considerations

Patient age and severity of CP were subpopulations to consider in the ACHIEVE study. Children who participated from lower GMFCS levels (higher function) and those with higher parent satisfaction made greater positive changes. Preterm birth, SES, and patient age were not associated with any changes. We had 12 stratification groups (2 [sex] × 3 [age group] × 3 [GMFCS levels] = 12). Within the total of 50 participants who were randomly assigned, there were fewer than 5 participants on average from each stratum. Although this randomization approach theoretically may ensure a balanced group assignment for each stratum, the small sample in each stratum may easily lead to imbalance. This is especially true when a large proportion of participants are randomized but not treated (ie, we had 20 participants who verbally consented to be randomized but were not able to participate in the study). More research is needed to identify the best approach to handle this issue in future studies. We did not check for treatment response to heterogeneity.

Study Limitations

This was a longitudinal study with a 9-month treatment period. As such, some of the changes could be attributed to natural history, and we did not have a no-treatment study group. The treatment for the weekly group in this study might not be the traditional standard of care because all treating therapists had training and were monitored for fidelity; in standard of care, treatment fidelity is not routinely monitored. The “active ingredients” of this type of treatment are efficacious. This study was designed originally to include only a randomized study. Due to low recruitment, patient-reported schedule challenges for families to participate, and ongoing feedback from a stakeholder advisory panel using some community-based participatory research approaches as part of the original study design, we added a self-selected cohort. We had a total of 100 participants at baseline; 80 finished the 9-month follow-up and 4 did not complete the GMFM assessment. The 12-month follow-up had 56 participants who completed the GMFM assessment.

The main limitations of this study were lack of power for the randomized study due to low recruitment and current attrition for follow-up. The current protocol included both cohorts (randomized and self-selected), with a total of 100 participants at baseline (46 randomized, 54 self-selected) and 76 participants with GMFM measures at the end of treatment (36 randomized and 40 self-selected). The statistical analyses considered both cohort and treatment (4 groups) in the model, along with the sensitivity analyses, including all available data assuming MAR or only those participants with completed treatment. The conclusions of the primary hypothesis from different analysis methods were consistent. In addition to the P values corresponding to the hypothesis testing, the 95% CIs of the treatment effect for each group and the differences between the 2 treatment groups were also reported to help readers interpret our results. Some of the cells for outcome measures, like the Bayley-III, which tested a subset of participants, had a much smaller sample size and were underpowered. Treatment for patients with CP is comprehensive. Although there were clear inclusion and exclusion criteria for participation in this study, we allowed other treatments but did not record them.