Explaining Variations in Quality of Mental Health Care for Children: Do

advertisement
Explaining Variations in Quality of Mental
Health Care for Children: Do
Organizational Characteristics of Clinics
Matter?
Coauthors: Susan E. Stockdale, Ph.D.1, Bonnie T. Zima, M.D.,
M.P.H.1, Michael S. Hurlburt, Ph.D.2, Penny Knapp, M.D.3
Supported by a grant from the California Department of Health Care Services
1
UCLA, Semel Institute Health Services Research Center; 2 Child and Adolescent
Services Research Center, Rady Children's Hospital, San Diego, CA; 3 California
Department of Mental Health
Background
• Zima, Hurlburt, et al. 2005 –
– Adherence to quality indicators in publicly-funded outpatient MH
care for children moderate to poor, varies considerably
depending on domain
• Hoagwood, Burns et al. 2001; Glisson and colleagues –
– Organizational structure, culture, capacity for QI associated with
sustainability, implementation of new services and programs
– Improvements in organizational infrastructure necessary
– Little is known about how social context of care interacts with
care processes
• Sociotechnical model –
– “Fit” between organizational context and technology important
for organizational effectiveness
Figure 1. Conceptual Model
External factors
-County median HH
income
- County population
density
Internal factors
Structure:
- ownership status
- number of services
- volume
-staffing
Climate/culture:
- hierarchy
- cooperation
Capacity for QI:
- case review
- outcome data
collection
- support for QI/QA
- medically-trained
staff
- use data for trt
planning/QI
Soft technology
- Quality domain:
- initial clinical assessment
- linkage to oth services
- basic trtment principles
- psychosocial treatment
- patient protection
- informed med decision
- general med monitoring
- med-specific monitoring
Child characteristics
- Gender
- Age
- Race/ethnicity
- Clinical severity
- Psychosocial complexity
Quality of care
-Probable
acceptable care
Research Objectives
• Main effects models
– Estimate the effects of child-level characteristics on
quality of care, with and without controlling for cliniclevel factors
– Explore the relationship between clinic-level factors
and quality of care, controlling for child-level
characteristics
• Interaction model
– Explore whether “fit” between soft technologies
associated with specific quality domains and cliniclevel characteristics explains variability in quality of
care
Data: Caring for California Initiative
(CCI)
• Child-level data: 813 children from 62 MH clinics in 21
CA counties
• Clinic-level data: interviews with 58 program managers
• Outcome: probable acceptable care for quality domain
(up to 8 indicators per child)
• Predictors
– Child-level: gender, race/ethnicity (white, black, Hispanic,
other/unknown), age (continuous), clinical severity (high vs low), and
psychosocial complexity (high vs. low)
– Clinic: ownership status (public vs private clinic), number of mental
health services offered, volume (number of children served annually),
staffing (percentage of clinicians available to treat children, cooperation,
hierarchy, measures of capacity to deliver quality care
•
Analysis: 3-level HLM with quality domain indicator, child, and clinic
levels.
Results of hierarchical logistic regression of child and
organizational predictors on probable acceptable care
OR (95% CI)
Model 1: child-level
only
Model 2: add clinic
level
Race/ethnicity (vs white)
Black
.95 (.73, 1.25)
.86 (.64, 1.17)
Hispanic
.86 (.68, 1.09)
.87 (.67, 1.13)
Other/unknown
.97 (.74, 1.27)
1.00 (.75, 1.33)
Age
.97 (.94, .99)*
.96 (.93, .99)***
Female
.90 (.74, 1.09)
.94 (.77, 1.15)
Clinical severity
1.28 (1.07, 1.54)**
1.33 (1.09, 1.62)***
Psychosocial complexity
1.08 (.90, 1.29)
1.07 (.89, 1.29)
County median HH income
=>$36,642
.94 (.73, 1.21)
.94 (.71, 1.23)
County pop density =>465/sq mile
NA
NA
Results of hierarchical models, cont.
OR (95% CI)
Model 1: child-level
only
Model 2: add clinic
level
Private clinic
------
NA
High pop density*private clinic
------
NA
No. MH services
------
1.14 (.85, 1.54)
% clinicians treat children
------
1.01 (1.00, 1.01)+
No. children served
------
1.00 (1.00, 1.00)
Cooperation
------
1.00 (.91, 1.09)
Hierarchy
------
.99
% cases reviewed by CQI/QA
------
1.00 (1.00, 1.01)
% collect performance data
------
1.00 (.99, 1.01)
No. QA supports
------
1.02 (.81, 1.28)
% medically trained FTE
------
1.01 (.99, 1.03)
Use data in treatment planning
------
1.15 (.84, 1.56)
(.93, 1.05)
Figure 1. Predicted probability of receiving acceptable care on the linkage domain
70.0%
61.9%
Percent receiving acceptable care
60.0%
50.0%
42.4%
39.5%
38.7%
40.0%
Directly operated clinics
Contract clinics
30.0%
20.0%
10.0%
0.0%
High
Low
County population density
Significant differences: private vs public in high pop; private vs.
public in low pop; private in high vs. low pop.
Figure 2. Predicted probabilty of receiving acceptable care on the basic treatment
principles domain
50.0%
45.0%
43.2%
40.0%
Percent receiving acceptable care
36.2%
33.9%
35.0%
28.6%
30.0%
Directly operated clinics
25.0%
Contract clinics
20.0%
15.0%
10.0%
5.0%
0.0%
High
Low
County population density
Significant differences: private vs public in low pop.
Figure 3. Predicted probability of receiving acceptable care on the psychosocial treatment
domain
100.0%
90.0%
87.3%
86.3%
Percent receiving acceptable care
80.0%
70.0%
74.9%
68.8%
60.0%
Directly operated clinics
50.0%
Contract clinics
40.0%
30.0%
20.0%
10.0%
0.0%
High
Low
County population density
Significant differences: private vs public in high pop.
Figure 4. Predicted probability of receiving acceptable care on the general medication
monitoring domain
90.0%
82.7%
80.0%
Percent receiving acceptable care
70.0%
61.4%
60.0%
50.9%
50.0%
45.7%
Directly operated clinics
Contract clinics
40.0%
30.0%
20.0%
10.0%
0.0%
High
Low
County population density
Significant differences: private vs public in high pop; private in high
vs. low pop.
Figure 5. Predicted probability of receiving acceptable care on the specific medication
monitoring domain
30.0%
26.7%
26.4%
Percent receiving acceptable care
25.0%
20.0%
17.1%
Directly operated clinics
15.0%
Contract clinics
10.0%
4.9%
5.0%
0.0%
High
Low
County population density
Significant differences: private vs public in high pop; public in high vs. low pop.
Conclusions
• Child age and clinical severity associated with quality
– younger, higher clinical severity – documented quality more
likely
• Clinic structure associated with quality
– Ownership status (private vs. public)
• External factors associated with quality
– County population density
• “Fit” – 3-way interaction between clinic structure,
population density, quality domain
– Differences likely due to unmeasured factors not captured by our
model
– Funding sources, differences in policies and regulations for
private vs. public clinics by county
Policy Implications
•
•
Performance on specific quality domains may take different
resources; linked with population density and ownership status
Linkage to other services
– low pop, lacking alternatives; high pop, more alternatives
•
Psychosocial treatments
– Private clinics in high population density counties - scored as probable
acceptable psychotherapy, even if not most appropriate for condition.
– Appropriateness of therapeutic modality, documentation of patient
outcomes should receive higher policy priority, technical assistance
where needed.
•
Medication monitoring
– Could reflect better availability of child psychiatrists in populated areas
– Private clinics are more competitive employers, not locked into county
pay scales
– Medication safety should be a policy priority
Download