Secondary Logo

Journal Logo


School-Based Physical Therapy Services: Predicting the Gap Between Ideal and Actual Embedded Services

Clevenger, Valerie D. PT, DSc, PCS; Jeffries, Lynn M. PT, DPT, PhD, PCS; Effgen, Susan K. PT, PhD, FAPTA; Chen, Sixia PhD; Arnold, Sandra H. PT, PhD

Author Information
doi: 10.1097/PEP.0000000000000683

The Individuals with Disabilities Education Act of 2004 (IDEA) mandates students receive their special education and related services in regular education environments (eg, classrooms, hallways, and playgrounds) unless students are unsuccessful in those environments with supports and services.1 In the educational environment, physical therapists, as related services providers, should embed their services within students' regular routines and activities, such as moving through classroom centers or climbing on playground equipment during recess.2 School-based physical therapists (SBPTs) acknowledge they do not always provide therapy services in regular or natural education environments.3 In a recent survey,4 SBPTs rated statements about their school-based physical therapy practice. Using a Likert scale to estimate their actual practice and thoughts about their ideal practice, 80% of SBPTs reported it was always or usually ideal to provide services in students' natural school environments. However, only 57% reported actually providing services in these settings always or usually, suggesting a gap existed between estimated ideal and actual practice for SBPTs that has remained consistent over 20 years.3,4

Research regarding the gap between estimated ideal and actual practice for physical therapy (PT) services embedded in students' routines and activities is limited. Family preference, therapy techniques or interventions, and individualized family service plan goals predicted the gap between estimated ideal and actual practice in early intervention,5 suggesting that these 3 variables may influence physical therapists' ability to embed services and may contribute to the gap between ideal and actual practice in early intervention. However, service delivery and team decision-making differ between early intervention and school-based services due to children's ages and team member composition.1 A predictive model for early intervention services cannot be applied directly to school-based PT services, leaving SBPTs without evidence about potential contributing variables such as family preferences, therapy techniques and intervention, and individualized student programs, that might improve their actual school-based practice.

Based on the IDEA,1 the gap could be predicted by the severity of students' disability if Individualized Education Program (IEP) teams are individualizing service location based on students' disability-related educational needs. Currently, no theories or predictive models exist to assist in decreasing the gap between SBPTs' estimated ideal and actual practice related to embedding PT services into students' natural school environment. Predictive modeling helps identify the relationships between variables, which could lead to new theories.6

Although research describing potential variables that could influence embedding therapy services within students' natural routines and activities is sparse, evidence suggests conflicting influences of additional potential variables. SBPTs reported that teachers' preference for, and parents' expectation of, pull-out therapy is barriers to embedding therapy services.7 However, SBPTs also reported that teachers' support is critical for embedded services8 and could influence embedding services. Thus, the influence from parents and teachers may either positively or negatively contribute to the estimated gap between ideal and actual practice in school-based services. In a national study, SBPTs reported embedded services were difficult to provide when their workloads were high because high workloads may constrain scheduling services during students' routines and activities (eg, during recess).7 However, SBPTs in the northeast United States had potentially lower workloads based on student-to-therapist ratios,9 and recommended fewer embedded services suggesting that lower workloads may be associated with fewer embedded services.10 Consequently, the influence of SBPTs' workload also is unclear.

While the aforementioned variables have some evidence, other possible variables are based on researchers' hypotheses. For example, SBPTs may be discouraged from embedding services so they can bill and be reimbursed from third-party payors.4 Districts may discourage embedding services by only reimbursing contracted SBPTs for direct services in isolated settings.3 Understanding the influence of these variables on the gap between ideal and actual practice may help inform SBPTs' practice.

The purposes of this study were to determine SBPTs' estimated difference between ideally and actually embedding school-based PT services and to ascertain the variables predicting the estimated difference between ideally and actually embedding school-based PT services. The 8 variables of interest include severity of students' disability (disability), PT interventions (interventions), students' IEP, goals (goals), inferred family preference (family), inferred teacher preference (teacher), SBPTs' workload (workload), ability to bill third-party payors (billing), and district written contracts that reimburse only for direct services (contracts).


Study Procedures and Participants

We used a nationwide online survey in our exploratory prospective study. Following the institutional review board approval, we shared the Web site link through the APTA Academy of Pediatric Physical Therapy (APPT) newsletter, the APPT school-based special interest group's e-mail list, and the PT COUNTS e-mail database.11 Using respondent-driven sampling,12 a form of snowball technique, participants/respondents were encouraged to invite other SBPTs to complete the survey via their own e-mail or the research Web site-generated e-mail. Inclusion criteria were licensed physical therapists, practicing at least 1 year in school-based PT, between the ages of 23 and 64 years, and currently practicing in a US school setting. Data collection occurred between August and November 2017.

Based on having 8 independent variables, we estimated needing at least 30 participants per variable to ensure reasonable predictive power of our model. Because we used a stepwise procedure, we increased the sample size to 50 participants per variable resulting in a needed sample size of 400 completed surveys.13


We developed an online survey based on a review of the literature regarding school-based practice. Online surveys can quickly reach a large number of participants in a short time.14 The survey included 57 questions divided into 3 parts: inclusion, demographic, and school-based service questions (see Supplemental Digital Content 1, available at: The school-based service questions addressed SBPTs' estimated ideal (n = 21) and actual (n = 21) percentage of services embedded in students' routines and activities based on 8 variables: (a) disability, (b) interventions, (c) goals, (d) family, (e) teacher, (f) workload, (g) billing, and (h) contracts. Embedded in students' routines and activities was operationally defined as providing services (intervention, treatment, etc) in the classrooms, gymnasiums, playgrounds, etc, as part of the students' typical activities (eg, participating in physical education class, playing at recess). We asked what percentage of students on SBPTs' caseloads actually received embedded PT services based on each variable. We then asked what percentage of students they thought should ideally receive embedded services based on each variable. To reduce response burden for SBPTs, answer choices for the school-based service questions were ordinal percentage ranges (0%-20%, 21%-41%, 41%-60%, 61%-80%, and 81%-100%). Actual practice questions included the additional choice of no such students on the participants' workload.

We examined content validity of the survey with 6 SBPTs and 1 early intervention physical therapist. Combined, these therapists had over 70 years of pediatric practice experience and 4 were enrolled in a pediatric doctor of science degree program. The 7 therapists independently took the survey and all agreed that the content reflected SBPTs' practice. Following therapists' recommendations, we changed wording to improve clarity and added 1 demographic question asking the grade-level range of SBPTs' workload. The survey was completed online using the Qualtrics15 survey system.

Data Analysis

The Qualtrics15 participant survey data were exported into Excel, checked for completeness, and found to be missing data in 8% of cells. To allow for full set analysis, we applied sequential multiple imputations, a procedure using statistics to infer best fitting answers to missing data cells.16 Regression imputation was used for ordinal and continuous variables and logistic imputation was used for nominal variables.16,17 We used descriptive statistics to characterize our participants without including imputed data.

The percentage of the 21 questions for which SBPTs answered 81% to 100% was calculated for each participant for the estimated ideal and actual school-based service questions and became the variables ideal and actual, respectively. The means of all participants' estimated ideal and actual percentages were computed to determine SBPTs' estimated difference between ideal and actual practice. We grouped the estimated actual percentages into 8 independent variable categories: disability (7 questions), interventions (2 questions), goals (2 questions), family (2 questions), teacher (2 questions), workload (2 questions), billing (2 questions), and contracts (2 questions). After comparing answers to ideal and actual question pairs (eg, ideal practice vs actual practice for students with mild disabilities), the percentage of questions in which the answer was greater for ideal than actual was calculated and referred to as the gap. Predictive modeling analysis was performed using SAS.18 Variables for the model were selected using stepwise linear regression procedures. The α level was 0.05.

We created 3 additional independent demographic variables: region, based on participants' state of residence using APPT regions (Region I West; Region II North Central; Region III Great Lakes; Region IV North East; Region V South East; Region VI South; Region VII South Central); advanced degree, based on comparing participants' entry level with their highest degree; and billing practices, based on whether or not participants reported having students on their workload for whom they could bill third-party payers. Demographic category pair means (eg, advanced degree vs no advanced degree) were analyzed with a paired t test for pairs with 2 categories and at least 30 participants per category; single-factor analyses of variance for demographic pairs were used with more than 2 categories or Kruskal-Wallis nonparametric tests when at least 1 category contained fewer than 30 participants. The α level was 0.05 and Bonferroni correction was used for multiple analyses.

For the school-based service question pairs (eg, more vs fewer hands-on interventions), χ2 statistics were used to compare the frequency of answering 81% to 100% for ideal and actual and having a difference for gap. The α was 0.05.


Of 727 therapists who opened the online survey, 410 participants completed 50% or more of the questions in each section. Participants represented the majority of the United States with the exception of Alabama, District of Columbia, and Puerto Rico. Most participants were older than 40 years (75.74%), APTA members (58.09%), APPT members (54.66%), employed by the district (65.36%), and working full time (79.85%). Although the majority of participants had over 20 years of experience as physical therapists (60.84%), the majority had 20 or fewer years of experience as SBPTs (71.57%) (Table 1).

TABLE 1 - Demographic Characteristics of SBPTs' Participants
Characteristics n (%) Mean (SD)
Advanced degree
Yes 161 (39.36)
No 248 (60.64)
Age, y
23-30 20 (4.90)
31-40 79 (19.36)
41-50 127 (31.13)
51-64 182 (44.61)
APTA member
Yes 237 (58.09)
No 171 (42.91)
APPT member
Yes 223 (54.66)
No 185 (45.34)
Billing practices
Yes 328 (80.79)
No 78 (19.21)
Employment status
Contracted 114 (28.01)
Employed 266 (65.36)
Other 27 (6.63)
Entry-level degree
BS 200 (48.90)
MPT 117 (28.61)
DPT 92 (22.49)
Highest degree earned
BS 104 (25.37)
MPT 54 (13.17)
DPT 194 (47.32)
MS 50 (12.20)
DSc 6 (1.46)
ScD 0 (0)
PhD 2 (0.49)
Other 0 (0)
I 33 (8.05)
II 57 (13.90)
III 93 (22.68)
IV 101 (24.63)
V 51 (12.44)
VI 27 (6.59)
VII 48 (11.71)
Work status
Full time 325 (79.85)
Part time 82 (20.15)
Years as PT
1-10 65 (16.01)
11-20 94 (23.15)
21-30 146 (35.96)
≥31 101 (24.88)
Years as SBPT
1-10 136 (33.33)
11-20 156 (38.24)
21-30 92 (22.55)
≥31 24 (5.88)
Workload, %
Prekindergarten 31.87 (26.48)
Elementary grades 42.06 (22.56)
Middle/junior high grades 12.51 (12.13)
High school grades 10.52 (12.07)
Post-high school/transition 2.34 (6.72)
Abbreviations: APPT, Academy of Pediatric Physical Therapy; BS, bachelor of Science; DPT, doctor of physical therapy; DSc, doctor of science; MPT, masters of physical therapy; MS, master's degree; PhD, doctor of philosophy; PT, physical therapist; SBPTs, school-based physical therapists; ScD, doctor of science; SD, standard deviation.
aRegion I: Alaska, California, Hawaii, Idaho, Montana, Nevada, Oregon, Utah, Washington; Region II: Colorado, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota, Wyoming; Region III: Illinois, Indiana, Michigan, Ohio, Wisconsin; Region IV: Connecticut, Delaware, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont; Region V: District of Columbia, Maryland, North Carolina, South Carolina, Virginia, West Virginia; Region VI: Alabama, Florida, Georgia, Kentucky, Mississippi, Puerto Rico, Tennessee; Region VII: Arkansas, Arizona, Louisiana, New Mexico, Oklahoma, Texas.

For SBPTs' estimated difference, the mean percentage of SBPTs who selected the 80% to 100% rating for ideal and actual was 50.02% and 32.56%, respectively. In our predictor model, disability, billing, contracts, and family predicted the gap between estimated ideal and actual practice (Table 2). A change in any 1 predictor variable resulted in a change in the gap. This model explained approximately 47% of the variability (R2 = 0.4718; adjusted R2 = 0.4666). No evidence of interaction or collinearity was found; however, the residuals were nonnormally distributed and heteroscedastic.

TABLE 2 - Significant Predictor Model Variablesa
Variable Parameter Estimate Standard Error t Value P Value
Disability −0.2258 0.0507 −4.45 <.0001
Billing −0.1459 0.0426 −3.42 0.007
Contracts −0.1391 0.0430 −3.24 0.001
Family −0.1280 0.0417 −3.07 0.002
R 2 0.4718
Adjusted R 2 0.4666
aThreshold for statistical significance set at P ≤ .05 and determined by stepwise linear regression.

Regarding participants' demographic analysis, estimated ideal practice was different for advanced degree, APTA membership, APPT membership, billing, employment status (contracted, employed, other) and region, specifically regions II and III (Table 3). Estimated actual practice was different for APTA membership, APPT membership, billing, employment status and region, specifically regions II and III, II and IV, III and V, and IV and V. The gap was different for billing, employment status, years as a physical therapist, and specifically regions II and III, II and IV, and IV and V.

TABLE 3 - Comparison of Demographic Pairs/Categories Based on Ideal, Actual, and Gap
Demographic Variable Ideal Actual Gap
Mean (Median)a P Value Mean (Median)a P Value Mean (Median)a P Value
Advanced degreeb
Yes 56.29 35.24 42.97
No 45.93 .0031 30.80 .0939 45.45 .2154
APTA membershipb
Yes 55.59 36.44 44.61
No 42.24 .0002 27.12 .0026 44.28 .4583
APPT membershipb
Yes 56.66 37.46 44.28
No 41.96 <.0001 26.59 .0005 44.71 .4449
Billing practicesb
Yes 48.48 31.20 45.60
No 56.59 .0470 38.34 .0483 39.68 .0498
Employment statusc
Contracted 45.99 (35.71) 27.78 (14.29) 48.58 (33.33)
Employed 49.64 (47.62) 33.19 (19.05) 43.35 (47.62)
Other 70.90 (95.24) .0055 46.38 (38.10) .0246 38.27 (28.57) .1842
Entry-level degreed
BS 53.64 34.92 41.46
MPT 49.41 32.36 46.03
DPT 42.91 .0798 27.64 .2317 49.07 .1324
Part-/full-time employmentb
Part-time 50.06 27.06 50.99
Full-time 50.01 .4964 33.93 .0515 42.84 .0236
I 53.25 (52.38) 38.82 (23.81) 38.96 (28.57)
II 61.65 (76.19) 47.54 (52.38) 31.66 (28.57)
III 44.60 (33.33) 23.35 (9.52) 49.87 (42.86)
IV 41.87 (33.33) 20.18 (9.52) 57.10 (51.14)
V 59.94 (66.67) 47.53 (42.86) 34.36 (23.81)
VI 47.62 (47.62) 36.16 (23.81) 38.27 (33.33)
VII 52.48 (52.38) .0089 36.41 (16.67) <.0001 40.67 (38.10) <.0001
Years as PTd
1-10 42.27 26.74 50.84 (57.14)
11-20 45.74 27.46 50.05 (47.62)
21-30 53.66 34.90 42.76 (38.10)
>30 53.59 .1049 37.54 .0720 37.77 (33.33) .0139
Years as SBPTsc
1-10 45.45 (33.33) 28.92 (14.29) 49.58 (47.62)
11-20 49.85 (47.62) 32.58 (19.05) 43.34 (38.10)
21-30 55.49 (59.52) 35.66 (23.81) 41.25 (33.33)
>30 56.15 (66.67) .2992 41.07 (33.33) .1893 35.32 (23.81) .0916
Significant pairs
Employment statuse
Contract—employed .3124 .1452 N/S
Contract—other .0035 .0420 N/S
Employed—other .0019 .0080 N/S
II-III N/S <.0001 .0020
II-IV .0011 <.0001 <.0001
III-V N/S .0004 N/S
IV-V N/S <.0001 <.0001
Years as PTg
11-20 to 31+ N/S N/S .0074
Abbreviations: APPT, Academy of Pediatric Physical Therapy; N/S, not significant; PT, physical therapist; SBPT, school-based physical therapist.
aMedian listed for nonparametric analysis only.
bThreshold for statistical significance set at .05 and determined by 2 sample t test assuming unequal variances.
cThreshold for statistical significance set at .05 and determined by Kruskal-Wallis nonparametric test.
dThreshold for statistical significance set at .05 and determined by analysis of variance.
eThreshold for statistical significance set at .0167 due to correction and determined by Wilcoxon rank sum test.
fThreshold for statistical significance set .00238 due to correction and determined by Wilcoxon rank sum test.
gThreshold for statistical significance set at .0083 due to correction and determined by Wilcoxon rank sum test.

When analyzing school-based service question pairs, we found the frequency of embedding services differed for estimated ideal for disability, family, goal, interventions, and teacher (Table 4). The frequency of embedding services differed for estimated actual for disability, family, goal, interventions, teacher, and workload. The frequency of having a gap between estimated ideal and actual practice differed for severity of disability.

TABLE 4 - School-Based Service Question Pair Comparisons
Demographic Variable Ideal Actual Gap
Numbera P Valueb Numbera P Value Numbera P Value
Able to bill 213 126 193
Unable to bill 220 .7314 127 .9577 195 .9219
High 196 115 198
Low 235 .0559 130 .4224 197 .9609
Direct only 208 111 225
Not direct only 216 .6951 151 .0357 193 .1171
More challenging 186 124 187
Less challenging 222 .0780 141 .3735 179 .6936
Mild 235 148 172
Mild-moderate 181 126 183
Moderate 160 115 190
Moderate-severe 174 138 163
Severe 205 <.0001 167 <.0001 141 .0046
Family preferences
Prefer isolated 142 82 179
Prefer embedded 265 <.0001 190 <.0001 158 .2961
Intervention based 149 97 180
Participation based 307 <.0001 201 <.0001 167 .5194
More hands-on 119 78 196
Less hands-on 240 <.0001 160 <.0001 173 .2577
Teacher preferences
Prefer isolated 168 97 189
Prefer embedded 266 <.0001 179 <.0001 171 .3746
aNumber/frequency of SBPTs who answered 80% to 100%.
bThreshold for statistical significance set at .05 and determined by χ2.


The results of our study suggest that the gap between SBPTs' estimates of ideally and actually embedding practice in school-based PT continues. Although SBPTs' estimates suggested that providing services during students' routines and activities is ideal, they estimated actually providing services to fewer students in those environments. Specifically, we found that a decrease of embedded services due to severity of students' disabilities, billing practices, SBPTs' employment contracts, and families' inferred preference, predicted an increase in the gap between embedded services SBPTs actually provide and the percentage SBPTs' estimate is ideal.

The strongest predictor in our model was disability. The frequency of SBPTs' estimates of providing embedded services and the frequency of SBPTs' estimated differences between ideal and actual practice varied between the severity levels of students' disabilities. These findings suggest that SBPTs may vary the amount of embedded services based on the nature and severity of the student's disability.1 Upon further analysis, SBPTs reported providing more services in natural environments for students with severe disabilities compared with students with less severe disabilities. Jeffries et al19 determined that SBPTs provided students with severe disabilities more individual and classroom-based services than students with less severe disabilities. SBPTs also provided more educational, positioning, and integumentary interventions to students with severe disabilities compared with students with less severe disabilities,19 which we hypothesize are more easily embedded than mobility-type interventions. Further, SBPTs reported the least amount of differences between estimated ideal and actual practice for students with severe disabilities, indicating SBPTs are embedding services at a level similar to what they consider ideal. Although not contemporary, some may consider the natural environment for some students with severe disabilities to be segregated classrooms or schools.20,21 We did not query the location of students' natural environments; thus, the location of students' natural environments warrants further exploration.

Interestingly, SBPTs provided the least amount of embedded services for students with moderate disabilities. We operationally defined moderate disability as including students who use assistive devices for walking. In our study, SBPTs had the highest gap score for students with moderate disabilities, indicating they were actually providing embedded services at a level less than they reported was ideal. We hypothesize that SBPTs may use interventions to address balance and strength that could be difficult to embed in students' routines and activities or implement interventions that support acquiring new skills (eg, learning to walk) that could be impractical to embed.22 SBPTs also may start interventions in less distracted environments before embedding interventions.23 Additional exploration is needed identifying why SBPTs' level of embedded services varies between different levels of students' disabilities and whether SBPTs embed services as students' skill proficiency increases.

Our second strongest variable was billing, an interesting finding considering that services are to be provided freely and appropriately to students.1 School systems, however, can bill Medicaid for medically necessary school-based services including physical therapy.24 Of the 47 states and Washington, District of Columbia, that have billing codes for school-based PT services, 92% of states used fee-for-service payment even if medical coverage was a managed care plan.24 Fee-for-service systems may encourage use of billable practices,25 which we hypothesize may occur in schools because the IDEA does not fully fund special education services.26 SBPTs may be encouraged to provide direct services in isolated settings or outside students' natural routines and activities, so schools can bill for therapy services. Because we did not ask SBPTs about payment systems, we do not know the influence of payment systems or the pressures that SBPTs may experience to provide billable services.

Our third strongest variable that predicated the gap between estimated ideal and actual practice was SBPTs' district written contracts. Interestingly, when we analyzed the difference between therapists' employment status, the gap between estimated ideal and actual practice did not differ between any of the employment groups. Therefore, although SBPTs' district written contracts predicted the gap, being a contracted or employed school personnel was not associated with the gap in practice.

Because administrators may lack understanding of the role of special education teachers,27 they may similarly not fully understand SBPTs' practice nor how to write contracts that support embedding services. For example, administrators may write SBPTs' contracts that allow high workloads leading SBPTs to provide isolated group interventions rather than embedded individual services. However, we did not find a significant difference in the gap between ideal and actual practice when comparing high and low workloads (Table 4). Further research may elucidate any relationship between SBPTs' district written contracts and use of embedded services.

Our survey only queried contracted SBPTs regarding district written contracts. These questions included the largest amount of imputed data of all the questions. Thus, the variable contracts should be interpreted with caution.

Our fourth predictor was inferred families' preferences. The frequency of SBPTs who embedded services varied based on whether therapists reported families' preferred embedded or isolated services. Specifically, more students received embedded services if their parents preferred services in regular education and fewer received embedded services if their parents preferred isolated services. Based on the results, SBPTs may take families' requests into consideration when making service decisions. Understanding parents' and students' goals and desires is important for SBPTs, as they strive to improve students' outcomes. Families may prefer isolated settings, particularly those accustomed to clinical services, or may not have considered contextual embedded services. As part of the IEP team, students, parents, and SBPTs should consult together to identify the context and location for effective service delivery. Although Thomason and Wilmarth7 reported that parents' expectations for isolated therapy services were barriers to providing services in students' natural settings, parents' opinions may differ on the emphasis of therapy,28 which could impact decisions regarding embedded services. While SBPTs may want to translate their knowledge of context-based, naturally occurring school-based therapy services to families, it will be important for SBPTs to acknowledge parents' preferences in service delivery.

Upon further analysis, we found regional differences in practice. SBPTs in the North region reported higher percentages of gap between estimated ideal and actual practice and fewer estimates of embedded actual practice compared with SBPTs in the North Central and South East regions. Similarly, Kaminker and colleagues10 found that northeastern therapists recommended fewer embedded services than other regions. They also found an increased number of independently contracted SBPTs in the northeast and suggested that contracted SBPTs may struggle more than employed SBPTs to be enmeshed within school systems to embed services.10 We hypothesize that regional differences may result from regional variance in the time students with disabilities spend in regular education classrooms,29 SBPTs' state practice guidelines,30 and potential student/therapist ratios.9 States educating fewer students with disabilities in regular education classrooms may influence embedding services if isolated services are considered acceptable because students spend more time isolated in special education settings compared with general education settings. Further research is needed to determine how regional factors may influence education and SBPTs' practice.

Finally, we found differences in SBPTs' estimated actual practice and gap based on demographic characteristics. Both APTA and APPT members embed more therapy services compared with nonmembers. We hypothesize members have more accessible education opportunities such as journals, newsletters, and networking on current and best practices. Interestingly, compared with recent school-based practice surveys,4,7,31 our survey had the largest sample of non-APTA members allowing for comparison to APTA members.

Future Research

Our study reveals relationships that could help develop a theory purporting why SBPTs do not provide services in students' natural environments at a level they consider ideal. However, we suggest a shift in research to determining predictors of actually embedding school-based services and on how embedding services may affect students' participation. Our study is the first to include variables of district written contracts and third-party billing which we suggest need continued evaluation and research. Further exploration of regions, and APTA and APPT membership, may elucidate the influence of these variables on SBPTs' practice.


Although we achieved our sample of over 400 participants, we excluded 36% of potential participants with incomplete surveys. Our survey was long, averaging 27 minutes to finish, which may have inhibited completion. Rather than asking families and teachers their preferences, we asked SBPTs to infer their preferences. Because SBPTs, as IEP team members, frequently collaborate with teachers and families, we think SBPTs have knowledge of teachers' and families' preferences for service delivery. While our model did not meet the assumptions of normality and homoscedasticity, our estimated regression coefficients and predictors are still unbiased due to our large sample size and central limit theory.32 Lastly, we used respondent-driven sampling12 to obtain a representative sample of SBPTs. However, SBPTs may have invited SBPTs with similar opinions and experiences. These biases may limit the generalizability of our study.


Although the gap between estimated ideal and actual practice continues, our predictive model is the first to specify 4 variables (disability, billing, contract, and family preferences) that predict the difference between what SBPTs think and what they actually do in providing PT services embedded within students' natural routines and activities. Further, this study is the first to report the analysis of SBPTs' estimates of actually providing embedded services and the gap in providing embedded services based on specific school-based predictor variables (severity of students' disability, PT interventions, students' IEP goals, inferred family preference, inferred teacher preference, SBPTs' workload, ability to bill third-party payors, and district written contracts) and APTA and APPT membership. Based on our results, further research is needed to explicate the influence of written contracts and billing practices in SBPTs' practice.


1. Individuals with Disabilities Education Act, 20 U. S. C. 2004.
2. Effgen SK, Chiarello L, Milbourne SA. Updated competencies for physical therapists working in schools. Pediatr Phys Ther. 2007;19(4):266–274.
3. Effgen SK, Klepper SE. Survey of physical therapy practice in educational settings. Pediatr Phys Ther. 1994;6(1):15–21.
4. Effgen SK, Kaminker MK. Nationwide survey of school-based physical therapy practice. Pediatr Phys Ther. 2014;26(4):394–403.
5. McWilliam RA, Bailey DB. Predictors of service-delivery models in center-based early intervention. Except Child. 1994;61:56–71.
6. Shmueli G. To explain or to predict? Stat Sci. 2010;25(3):289–310.
7. Thomason HK, Wilmarth MA. Provision of school-based physical therapy services: a survey of current practice patterns. Pediatr Phys Ther. 2015;27(2):161–169.
8. Sekerak D, Kirkpatrick DB, Nelson KC, Propes JH. Physical therapy in preschool classrooms: successful integration of therapy into classroom routines. Pediatr Phys Ther. 2003;15:93–104.
9. Effgen SK, Myers CT, Myers D. National distribution of physical and occupational therapists serving children with disabilities in educational environments. Phys Disabil. 2007;26:47–61.
10. Kaminker MK, Chiarello LA, Smith JAC. Decision making for physical therapy service delivery in schools: a nationwide analysis by geographical region. Pediatr Phys Ther. 2006;18(3):204–213.
11. Effgen SK, McCoy SW, Chiarello LA, Jeffries LM, Bush H. Physical therapy-related child outcomes in school: an example of practice-based evidence methodology. Pediatr Phys Ther. 2016;28(1):47–56.
12. Handcock MS. Discussion of “Adaptive and ne2rk sampling for inference and interventions in changing populations” by Steven K. Thompson. J Surv Stat Methodol. 2017;5(1):29–33.
13. VanVoorhis CRW, Morgan BL. Understanding power and rules of thumb for determining sample sizes. TQMP. 2007;3(2):43–50.
14. Dillman DA, Smyth JD, Christian LM. Internet, ph1, Mail, and Mixed-Mode Surveys: The Tailored Design Method. 4th ed. Hoboken NJ: John Wiley & Sons; 2014.
15. Qualtrics home page. Accessed August 2018.
16. Raghunathan TE, Lepkowski J, Van Hoewyk JH, Solenberger PW. A multivariate technique for multiply imputing missing values using a sequence of regression models. Surv Methodol. 2001;27(1):85–95.
17. Zhu J, Raghunathan TE. Convergence properties of a sequential regression multiple imputation algorithm. J Am Stat Assoc. 2015;110(511):1112–1124.
18. SAS [computer program]. Version 9.4. Cary, NC: SAS Institute.
19. Jeffries LM, McCoy SW, Effgen SK, Chiarello LA, Villasante Tezanos AG. Description of the services, activities, and interventions within school-based physical therapist practices across the United States. Phys Ther. 2019;99(1):98–108.
20. Kleinert H, Towles-Reeves E, Quenemoen R, et al. Where students with the most significant cognitive disabilities are taught: implications for general curriculum access. Except Child. 2015;81(3):312–328.
21. Kurth JA, Morningstar ME, Kozleski EB. The persistence of highly restrictive special education placements for students with low-incidence disabilities. Res Pract Persons Severe Disabil. 2014;39(3):227–239.
22. Palisano RJ, Murr S. Intensity of therapy services: what are the considerations? Phys Occup Ther Pediatr. 2009;29(2):107–122.
23. Kenyon LK, Blackinton MT. Applying motor-control theory to physical therapy practice: a case report. Physiother Can. 2011;64(3):345–354.
24. Baller JB, Barry CB. State variation in school-based disability services financed by Medicaid. J Disbil Policy Stud. 2016;37(3):148–147.
25. Mandell DS, Machefsky A, Rubin D, Feudtner C, Pita S, Rosenbaum S. Medicaid's role in financing health care for children with behavioral health care needs in the special education system: implications of the Deficit Reduction Act. J Sch Health. 2008;78(10):532–538.
26. Katsiyannis A, Yell ML, Bradley R. Reflections on the 25th anniversary of the Individuals with Disabilities Education Act. Remedial Spec Educ. 2001;22(6):324–334.
27. Roberts CA, Ruppar AL, Olson AJ. Perceptions matter: Administrators' vision of instruction for students with severe disabilities. Res Pract Persons Severe Disabil. 2018;43(1):3–19.
28. LaForme Fiss AC, McCoy SW, Chiarello LA. Comparison of family and therapist perceptions of physical and occupational therapy services provided to young children with cerebral palsy. Phys Occup Ther Pediatr. 2012;32(2):210–226.
29. US Department of Education. Thirty-ninth Annual Report to Congress on the Implementation of the Individuals with Disabilities Education Act, 2017.
30. Vialu C, Doyle M. Determining need for school-based physical therapy under IDEA: commonalities across practice guidelines. Pediatr Phys Ther. 2017;29(4):350–355.
31. Kaminker MK, Chiarello LA, O'Neil ME, Dichter CG. Decision making for physical therapy service delivery in schools: a nationwide survey of pediatric physical therapists. Phys Ther. 2004;84(10):919–933.
32. Long JS, Ervin LH. Using heteroscedasticity consistent standard errors in the linear regression model. Am Stat. 2000;54(3):217–224.

pediatric physical therapy; physical therapy; related services; school-based physical therapy

Supplemental Digital Content

© 2020 Academy of Pediatric Physical Therapy of the American Physical Therapy Association