1. Modeling Mental Health Recovery
Using a Hierarchical Linear Growth Model
Karen Traxler, M.S., Suzy Landram, M.S., Tyler Kincaid, M.S., & Lisa Rue, Ph.D.
Applied Statistics and Research Methods
University of Northern Colorado
ASA: Women In Statistics; Raleigh Durham, North Carolina: May 15-17, 2014
Abstract Results & DiscussionMethods
Purpose
Participants
All participants were residents of CooperRiis Healing Farm in North Carolina
Demographics Waves of Data
Measuring Recovery
• The Mental Health Recovery Measure-Revised (Young and Ensing, 2003) is a 30-item self-report survey used to
measure recovery outcomes across five domains: overcoming stuckness , self-empowerment, learning and self-
redefinition, basic functioning, and overall wellbeing. A five-point traditional Likert scale is used to measure
responses from 1= strongly disagree to 5= strongly agree. Higher scores indicate more of the trait of recovery.
The MHRM-R is also used as a unidimensional measure of recovery where the scores are aggregated across all
30 items and can range between 30 and 150. Cronbach’s alpha for the global scale = .941 and RMSEA = .084.
Data Analysis: 5 waves of data
• Using HLM 7 and SPSS (PASW, 22.0, 2013), only cases with no missing data, a two-level Hierarchical Linear Growth
Curve Model (HLM) was used to assess mental health recovery over time. Five time periods (or waves of data)
were used: admission, 3 months, 6 months, 9 months, and 12 months.
• The strength of a HLM Growth Curve Model over other regression models is that it can differentiate between
individual starting points (intercepts) as well as individual change/recovery over time (slopes) (Raudenbush &
Bryk, 2002).
Procedures
• Grand mean centering was applied to an individual’s MHRM-R scores at each wave point to simplify the
interpretation of scores
• Maximum Likelihood Estimation (MLE) was used for all analysis
Building the Model
Level-1 Model: The unconditional Model
MHRMRti = π0i + π1i*(TIMEti) + eti
Level-2 Model: The Conditional Model
π0i = β00 + β01*(ANXIETYi) + β02*(BIPOLARi) + β03*(DEPRESSIi) + β04*(SCHIZOPHi)
+ β05*(PERSONALi) + β06*(AGEi) + β07*(GENDERi) + r0i
π1i = β10 + r1i
Based on the RCC approach to mental health recovery, the purpose
of this study was to:
(a) Assess whether mental health recovery outcomes, based on the
MHRM-R improved over time for the target population
(b) Investigate the roles age, gender, and/or primary diagnosis play
in positive mental health recovery outcomes over time
Target Population: Individuals with Severe and Persistent
Mental Health Challenges
Statistically modeling mental health recovery for individuals with severe
and persistent mental health challenges has traditionally been
accomplished using the medical model of recovery which encompasses
the elimination or reduction of symptoms through medication and/or
hospitalization. A more holistic approach gaining support among both
clinicians and consumers is the Recovery-Centered Collaborative
Approach to mental health recovery (the RCC model) which is a
person-centered approach integrating medication with spirituality,
hope, physical wellbeing, life skills, strategies for managing symptoms,
and strong community and family support. Methods: Data from
CooperRiis Healing Farm, a residential treatment facility for individuals
with severe and persistent mental health conditions, specializing in the
RCC approach were examined using a Hierarchical Linear Growth
Model (HLGM) to assess recovery over a twelve month time period.
Data included variables such as age upon admission, gender, and
primary diagnosis. Results: The results from level 1 of the HLGM
provided evidence that there was significant positive growth in
recovery scores over time. Results from level 2 of the HLGM revealed
that individuals with a diagnosis of personality disorder or depression
had admission scores significantly lower than any other diagnoses.
Variance in growth over time was not explained by any of the level 2
variables
Special thanks to:
Dr. Sharon Young &
Matt Snyder, M.S., M.A., L.P.C.
CooperRiis Healing Community
Special thanks to:
Dr. Susan Hutchinson
University of Northern Colorado
Research Questions
• Interactions: Interactions of independent variables were tested and no significance
was found
• Proportion of Variance Explained: The primary diagnosis of the residents explains
35.3% of the parameter variance in the initial status (i.e., where a given resident
baseline recovery score will start) Future Research: Variance in growth over time has
not been explained by any level 2 variables in the model.
• Personality and Depression
Scores over time
Implications for Mental Health Researchers, Clinicians, and Consumers
• This study supports the holistic recovery-centered collaborative (RCC_ approach to
recovery as a viable alternative to the medical model of mental health recovery,
even for patients with the most severe and persistent mental health disorders
• Applied researchers and clinicians can use this information to develop appropriate
person-centered treatments for severe mental health conditions
• Individuals seeking mental health treatment for acute symptoms can have a voice in
their recovery process and maintain hope throughout their journey of symptom
management.
Limitations
Recovery is a complex process involving far more than age, gender, and primary
diagnosis, therefore significant limitations are inherent in any explanatory model of
recovery over time
• Research Question # 1: Does mental health recovery of individuals with
severe and persistent mental health conditions, receiving a recovery-
centered collaborative mental health intervention, improve over time?
o Is there a difference in recovery outcomes over time based on:
• Research Question # 2: gender?
• Research Question # 3: primary diagnosis?
• Research Question # 4: age?
All statistical tests were conducted with α =.05.
• Research Question # 1: Evidence supported growth (recovery) over time
Level 1: Unconditional Model:
This model only included the MHRM-R recovery scores as the outcome variable and TIME as an independent variable.
The unconditional model determined that there was indeed growth over time, allowing the addition of Level 2 to the
model, where possible explanatory variables were included (Raudenbush & Bryk, 2002).
Level 1 Intercepts: Estimations of the mean (pooled) intercepts, 𝛽00= 105.88, p < .0001, were significant indicating
significant differences in scores on the MHRM-R at admission (baseline)
Level 1 Mean Growth Trajectories: The mean growth rate, 𝛽10=21.21, p < .0001, for the MHRM-R recovery scores was
significant, providing evidence that residents were gaining an average of 21.21 points over the 12 months of recorded
MHRM-R recovery scores
(1)
Level 2: Conditional Model: See Table 3
Findings from the unconditional model confirmed the requisite of a conditional model
with explanatory variables of personal characteristics (i.e., age, gender, and primary
diagnosis) being added to the model, at Level 2:
• Research Question # 2: There was not a significant difference in the growth
trajectory of recovery outcomes based on gender (𝐵11 = −1.49, 𝑝 = .724).
• Research Question # 3: Baseline Scores: Scores on the MHRM-R differed
significantly based on primary diagnosis; individuals presenting with depression
(𝐵01 = −9.51, 𝑝 = .026) or personality disorders (𝐵02 = −9.58 , 𝑝 = .046) scored
significantly lower upon admission than those with other diagnoses regardless of
gender or age of individual
• Research Question # 4: :There was not a significant difference in recovery
outcomes based on the residents’ age for the intercept (i.e., where the residents’
started; p = 0.620, nor in their growth trajectories (i.e., the residents’ recovery over
time; p = 0.358)
Implications & Limitations
(2)
Personality Disorders in red Scores upon admission differed
significantly based on primary diagnosis
Table 1
Years N Gender Primary Diagnosis n
2003-2013 285 Anxiety 5 Range 18-71
Bipolar 13 Mean 31.17
Depression 6 SD 11.60
Schizophrenia 97
Personality Disorder 23
Anxiety 12
Bipolar 20
Depression 8
Schizophrenia 39
Personality Disorder 15
Participants' Descriptive Statistics
Age (in years)
Male
Female
The MHRM-R is a self report
measure, and while results show
significant recovery scores, they
are based on the treatment
received at CooperRiis and may
have only moderate external
validity
[a] [b]
The growth model was based on 12
months of data. Trajectories may or may
not have improved with additional waves
[c]
Some individuals completed the
MHRM-R up to one month following
admission and, therefore, experienced
the benefits of treatment prior to their
first assessment
Depression
in red
Results & Discussion
Table 2
Wave of Data Time Period N
1 Admission 285
2 3 months 285
3 6 months 285
4 9 months 285
5 12 months 285
Time Periods of Data
Collection with Corresponding
Sample Sizes
Table 3
Standard Approx.
error d.f.
INTRCPT2, β 00 105.8855 1.795575 58.970 281 <0.001
DEPRESSION, β01 -9.5800 2.154347 0.851 281 0.046
PERSONALITY, β02 -9.5100 3.2144 -1.631 281 0.026
AGE, β 03 -0.0913 0.091555 -0.997 281 0.319
GENDER, β 04 -1.4119 2.042749 -0.691 281 0.490
INTRCPT2, β 10 21.21 2.277318 9.316 284 <0.001
Final Estimation of Fixed Effects in the HLM Model
Fixed Effect Coefficient t -ratio p-value
For INTRCPT1, π 0
For TIME slope, π 1