Post on 18-Jan-2016
Multifaceted intervention to improve health worker adherence to Integrated
Management of Childhood Illness (IMCI) guidelines in Benin*
* Published in American Journal of Public Health 2009;99:837–846.
ICIUM_P6a_presentation_v3.pptx
Alexander K. Rowe, Faustin Onikpo, Marcel Lama, Dawn M. Osterholt,
Samantha Y. Rowe, Michael S. Deming
Background and objective
• WHO’s IMCI strategy aims to improve child health in developing countries by encouraging use of evidence-based guidelines for treating major causes of death
• Studies show that training health workers (HWs) on IMCI guidelines can improve quality of care at outpatient health facilities (HFs)
• However, these studies also revealed substantial room for improvement in adherence to guidelines—even after IMCI training
• Objective: Evaluate an intervention to support HWs after IMCI training in Benin
Methods• Setting: 130 public and licensed private outpatient HFs
• Design: Randomized trial in 16 districts− Two similar areas, 8 districts each; one area randomly
selected as intervention area− HWs in both areas received IMCI training− HWs in intervention area also received package of post-
IMCI-training supports (next slide)− HWs in control area received “usual” supports
(i.e., whatever government provided) − Baseline: In 1999, assessed health care quality before
IMCI training with HF survey (observed consultations, re-examined patients, and interviewed caretakers & HWs)
− Eleven-day IMCI courses took place from 2001–2004− Follow-up HF surveys in 2001, 2002, and 2004
Methods
• Study population: Ill children <5 years at HFs and HWs
• Outcome measures: 1) Children with a potentially life-threatening illness (PLTI) with
recommended treatment (exactly according to IMCI)
2) Children with PLTI with recommended or adequate treatment (not exactly according to IMCI, but still considered life-saving)
3) For all children, an index of overall guideline adherence (% of all IMCI-recommended tasks that were performed)
• Package of post-training supports for IMCI-trained HWs1) Job aids: IMCI patient register and counseling guide
2) Supervision of HWs and supervisors
3) Nonfinancial incentives: Framed certificate of merit presented at a ceremony
Analysis
• Effect size = absolute %-point difference of differences: Effect size = (%POST – %PRE)intervention – (%POST – %PRE)control
• Intention-to-treat analysis: Regression models with time, study area, and time x area interaction terms
• Per protocol analysis:− Needed because IMCI training occurred slowly, and poor
quality care delivered by non-IMCI-trained HWs diluted performance in both study arms
− HWs divided into three groups: 1) HWs with IMCI training + study supports
2) HWs with IMCI training + usual supports
3) HWs without IMCI training
− All F/U surveys (2001–2004) pooled, as results were similar− Model: time, HW group, and time x group interaction terms
Effect
Results: Enrollment & intention-to-treat analysis
• We observed 1244 ill child consultations; 1101 of these were for PLTIs
• In F/U period (2001–2004), although we expected all children to be seen by IMCI-trained HWs, only half were− Partly because IMCI training occurred slowly− Partly because IMCI-trained HWs were not always
scheduled to work during peak hours
• Intention-to-treat analysis: All outcomes improved over time; however, differences in improvements between intervention and control areas were close to zero and not statistically significant (next slide)
Intention-to-treat analysis
0
20
40
60
80
100
Intervention area
40%
35%22%
16%
25%
32%Control area
1999 2001 2004
Year of survey
Per
cent
of
child
ren
rece
ivin
g re
com
men
ded
care
Baseline (pre-IMCI)
Follow-up
0
20
40
60
80
100
IMCI + “usual” supports
Not IMCI trained
IMCI + study supports
1999 2001 2002 2004
Per
cent
of c
hild
ren
rece
ivin
g re
com
men
ded
care
Baseline (pre-IMCI)
Follow-up
Per protocol analysis: Outcome 1 (recommended tx)
Graph shows performance trends for 3 HW groups. Trends for the other 2 outcomes were very similar.
0
20
40
60
80
100
IMCI + “usual” supports
Not IMCI trained
IMCI + study supports
1999 2001 2002 2004
Per
cent
of c
hild
ren
rece
ivin
g re
com
men
ded
care
Baseline (pre-IMCI)
Follow-up
Per protocol analysis: Outcome 1 (recommended tx)
For all outcomes, HWs without IMCI performed poorly, and quality did not change over time.
0
20
40
60
80
100
IMCI + “usual” supports
Not IMCI trained
IMCI + study supports
1999 2001 2002 2004
Per
cent
of c
hild
ren
rece
ivin
g re
com
men
ded
care
Baseline (pre-IMCI)
Follow-up
Per protocol analysis: Outcome 1 (recommended tx)
For all outcomes, IMCI-trained HWs with usual supports out-performed HWs without IMCI training by 19–35 %-points (p<0.05).
0
20
40
60
80
100
IMCI + “usual” supports
Not IMCI trained
IMCI + study supports
1999 2001 2002 2004
Per
cent
of c
hild
ren
rece
ivin
g re
com
men
ded
care
Baseline (pre-IMCI)
Follow-up
Per protocol analysis: Outcome 1 (recommended tx)
IMCI-trained HWs with study supports outperformed IMCI-trained HWs with usual supports by 15–27 %-points; statistically significant for 2 of 3 outcomes.
Potential limitations
• Study design changed from pre-post study with randomized controls (intention-to-treat) to pre-post study with non-randomized controls (per protocol analysis)
• As it is highly unlikely that child’s caretaker would know if HWs received study or usual supports, the change in design was probably not an important source of bias
• Other potential limitations
Bias introduced by observing consultations
Some prescriptions were incomplete
It is difficult to know if “usual” supports in our study reflect usual supports elsewhere
Conclusions
• Training, such as IMCI training, can be useful for implementing clinical guidelines; but it is not enough
• Relatively inexpensive post-training supports, such as those in our study, can lead to additional improvements
• However, even with our study supports, there was considerable room for improvement
• Wherever IMCI is implemented, program managers should ensure that IMCI-trained health workers perform consultations
Thank you!
Merci !