Upload
roger-whitehead
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
Hopkins PSMG Oct 2005 1
An Incomplete Model for Trial Designs for Moving to Scale:
The Whole Day ProgramC Hendricks BrownSheppard KellamJeanne PoduskaAmy Windham
John ReidCarla Ford
Natalie Keegan
Supported by R01 DA15409-02 PI: KellamRo1 MH40859-17 PI: Brown
Baltimore City Public School System
Hopkins PSMG Oct 2005 2
Premise
The Whole Day Program is where the school would want to be in 5 years, systemwide.
What specific steps should be examined along the way?
Hopkins PSMG Oct 2005 3
Primary Question
Effective?Who benefits, for how long?Who doesn’t benefit, who is harmed?
Efficacy Trial12 Schools24 Classrooms600 Students3 Time Periods
Hopkins PSMG Oct 2005 4
Differential impact on spring of first grade reading achievement by
baseline reading (fall of first grade): control against two interventions
Control
Classroom / Parent Interventions
Brown: Statistical Power
Hopkins PSMG Oct 2005 5
Presuming Effectiveness
SustainabililtyDo WD teachers continue to sustain the same
effects on children that they have in following years?
12 WD Teachers vs Themselves & 12 Wait-Listed Controls
Scalability Can program be successfully implemented
outside the original 12 schools throughout the school system?
Hopkins PSMG Oct 2005 6
Scale to What?
• From 2 To all 4 of the School Areas
• From 12 To 20 To 135 Schools
• From 24 To 40 To 400 First-Grade Classrooms
Hopkins PSMG Oct 2005 7
An Evidence-Based Decision Tree
Efficacious especially w/ High-Risk Youth?
Sustainable?
Feasible to Scale?
Y
Y
Go to Scale!
Y
N
N
N
Hopkins PSMG Oct 2005 8
The Real Decision Tree for BCPSS
Something Different
Hopkins PSMG Oct 2005 9
Person, Place, and Time
• Child• Classroom / Teacher (1st Grade)• School• Area (Region of Baltimore)
• Time – Scale(s): 10 Secs, Min, Quarter, Year• Child : Track from 1st through 3rd Grade• Teacher: Track over 3 years (cohorts)
Hopkins PSMG Oct 2005 10
Randomization
• 20 Schools in 2 Areas randomly assigned to One of 12 schools where trial begins
One of 8 schools that are “wait-listed”
• Children randomly assigned to classes within school
• Teachers/Classes randomly assigned to WD or wait-list
Hopkins PSMG Oct 2005 11
Baltimore Whole Day TrialAssignment of Classrooms/Teachers in First YearHalf the Teachers are Wait-Listed for WD to 3rd
Year
Replicate Same Design
w/ SC and and WD
BOTH in 12 Different Schools
Brown: Statistical Power
Hopkins PSMG Oct 2005 12
Effectiveness Evaluation
Compare 3 years of Outcomes for Children in the 12 WD classes to those in 12 Standard classes -- 1st Cohort
Examine Level of Program Implementation in the 12 WD and 12 Standard Classrooms in 1st Cohort
Hopkins PSMG Oct 2005 13
Classroom Observations by Independent Observers
3 times per year WD and SC classroomsAM-Reading InstructionPM-Other subjects
Teacher PracticesTimed ObservationsChecklist Global Ratings of Quality
Student BehaviorOn-Task/Off-taskRatings of Disruptive, Aggressive, and Shy Behavior
Hopkins PSMG Oct 2005 14
Classroom Observation Data Collection FormTeacher Practices and Student Behavior
Actual Form
Hopkins PSMG Oct 2005 15
Domains of Reading Instruction
• Word Work – phonemic awareness, alphabetic instruction, letter recognition, work work
• Reading Comprehension – reading comprehension, previewing a book, reading books
• Language – oral language, vocabulary, reading own writing, writing, grammar
• Non-reading Instruction – interruptions, teacher out of classroom
• Giving instructionally related reading directions• Feedback in response to behavior• Feedback in response to a reading relevant
comment or event
Hopkins PSMG Oct 2005 16
Classroom Observations BaselineData Overview
12 schools24 teachers (12 Whole Day, 12 Standard setting)
Each teacher observed 3 times per year, at each time point:
– 1 reading instruction (am observation)– 1 non-reading instruction (pm observation)
Length of observations at each time point:
Teacher: mean = 164 minutes Each student: mean = 9 min/student
Hopkins PSMG Oct 2005 17
Given Teacher Instructional Practices, Classroom Behavior(Standard Practice Classrooms during Reading Instruction at
Baseline)Well behaved v Fairly well behaved: X2=36.83 p<10-7
Word Work Reading Comprehension
Language
Well behaved 216
67.3%
184
61.3%
84
39.6%
Fairly well behaved
99
30.8%
109
36.3%
115
54.2%
Poorly behaved
1
.3%
7
2.3%
13
6.1%
Chaotic 5
1.6%
0 0
Hopkins PSMG Oct 2005 18
Student behavior: Reading v Other subjects(Standard Practice Classrooms at Baseline)
Never off-task v Ever off-task: X2=29.80 p<10-8
AM-Reading PM-Other subjects
Never off task 1138
95%
679
88.3%
Off task-1 interval 13
1.1%
13
1.7%
Off task-2 intervals 5
.4%
10
1.3%
Off task-3 intervals 42
3.5%
67
8.7%
Hopkins PSMG Oct 2005 19
Effectiveness Results for Cohort 1 (1 of 2)
• At Baseline Level of Off-Task Behavior is Well Balanced across Intervention Condition
• Reduction in Off-task BehaviorControl 21% WD 9%
• 50% Reduction in Aggressive Behavior for Males in Less Structured Settings
Hopkins PSMG Oct 2005 20
Graphical Comparison for Off-Task Behavior
Q-Q Plot of Males Off-Task Behaviorin Afternoons at End of First Grade
Global Rating for Control Classrooms
Glo
ba
l Ra
ting
fo
r W
D C
lass
roo
ms
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Hopkins PSMG Oct 2005 21
Comparison of Teachers
Q-Q Plot of Teacher Global Overall RatingEnd of First Grade
Global Rating for Control Classrooms
Glo
ba
l Ra
ting
fo
r W
D C
lass
roo
ms
1 2 3 4 5
3.0
3.5
4.0
4.5
5.0
Hopkins PSMG Oct 2005 22
Unsolved Multilevel Modeling Issues
Child Gender, Baseline Off-TaskClassroom Intervention, Classroom
Baseline AggrSchool
Time-Varying Classroom Context
Binary Outcomes
Hopkins PSMG Oct 2005 23
Sustainability Evaluation (Funded)
Multilevel Support
Compare Teacher Implementation (and Child Outcomes) for 12 WD Teachers in 1st through 3rdYear for First Grade Classes
Hopkins PSMG Oct 2005 24
Sustainability Hypotheses and N’s & Years
• Teacher practices over time will be related to the quality of WD integration at the level of the School Building Team.
12 20 135
2• Teacher practices over time will be related to
the quality of WD integration at the level of the School Area over time.
2 4
Hopkins PSMG Oct 2005 25
Scalability Design
Change in fidelity (Teacher practices) and impact (Child Outcomes) in succeeding years after initial training
12 24 135
Hopkins PSMG Oct 2005 26
Combined RCT and Implementation Trial
Trial After a Trial Design
Full Implementation Control Waiting
Effectiveness Trial
Yr1 Yr2
Yr3
Moving to Scale
Full Implementation Control
Varying Levels of Implementation Due to Sustainability and Scalability
Hopkins PSMG Oct 2005 27
How to Go From 24 to 400 Classes?
A Dynamic Wait-Listed type of design
Randomly Assign WHEN schools and classes obtain training, coaching, and full implementationContinue to assess intervention fidelityPaired-down measure of Teacher Practices
Random Assign Children to Classes?
Hopkins PSMG Oct 2005 28
ClassesSchools
Waiting
Active School
Trained
Wait Listed
Sustained
Trained Sustained
Hopkins PSMG Oct 2005 29
Plausible Time Scale # Classes (Schools)
Time Period/Cohort
New WD Sustainability Comparison Wait-Listed
Effectiveness 12 (12) 0 12 (Same 12) 376 (123)
Sustain 0 12 (12) 12 (Same 12) 376 (123)
Scalability 12 (12) 12 (Same 12) 20 (8) 356 (115)
Full Scalability
Yr 1
90 (30) 44 (20) 40 (30) 230 (85)
Yr 2 90 (25) 130 (50) 40 (25) 140 (60)
Yr 3 90 (30) 220 (70) 40 (30) 50 (35)
Yr 4 90 (35) 310 (100) 0 (0) 0 (0)