)DFWRUVFRQVWUDLQLQJDGKHUHQFHWR
UHIHUUDODGYLFHIRUVHYHUHO\LOOFKLOGUHQ
PDQDJHGE\WKH,QWHJUDWHG0DQDJHPHQW
RI&KLOGKRRG,OOQHVVDSSURDFKLQ
,PEDEXUD3URYLQFH(FXDGRU

HD Kalter1, R Salgado2, LH Moulton3, P Nieto4, A Contreras2, ML Egas4, RE Black1

Department of International Health, Johns Hopkins University Bloomberg School of Public Health,
Baltimore, MD, USA;

The BASICS Project, Rosslyn, VA, USA;

Departments of International Health and Biostatistics, Johns Hopkins University Bloomberg School of
Public Health, Baltimore, MD, USA;

Universidad Tcnica del Norte, Ibarra, Ecuador.

R Salgado is currently with John Snow, Inc., Arlington, VA, USA.

Short title: Referral constraining factors for severely ill children

Corresponding author: H.D. Kalter, Department of International Health, Johns Hopkins University Bloomberg School of
Public Health, 615 North Wolfe Street, Room E8132, Baltimore, MD, 21205, USA. E-mail hkalter@jhsph.edu. Tel +1
410 955 3928. Fax +1 410 614 1419.

&RQWULEXWRUV
HD Kalter designed and helped develop the study, helped analyze the data, and wrote the paper.
R Salgado helped design and develop the study, and contributed to the writing of the paper.
LH Moulton conducted the statistical analysis and contributed to the writing of the paper.
P Nieto helped develop the study and analyze the data, managed the project, and contributed to the writing
of the paper.
A Contreras helped develop the study and analyze the data, and contributed to the writing of the paper.
ML Egas supervised the project and contributed to the writing of the paper.
RE Black helped develop the study and contributed to the writing of the paper.

$EVWUDFW

Kalter HD, Salgado R, Moulton LH, Nieto P, Contreras A, Egas ML, Black RE. Factors constraining
adherence to referral advice for severely ill children managed by the Integrated Management of
Childhood Illness approach in Imbabura Province, Ecuador. Acta Paediatr 00, Stockholm. ISSN 08035233
Aim: Low referral completion rates in developing countries undermine the Integrated Management of
Childhood Illness (IMCI) strategy for lowering child mortality. We sought to identify factors
constraining adherence to referral advice in a health system using the IMCI approach.
Methods: We prospectively interviewed caregivers of 160 children urgently referred to hospital.
Caregivers who accessed and did not access hospital were compared for potential referral
constraining factors, including demographics, family dynamics, the severity of their childs illness, their
interaction with the health system, self-perceived problems, and physical and financial access.
Results: 67/160 (42%) referred children did not access hospital. 6 factors were associated with nonaccess, including 2 health worker actions: not being given a referral slip (adjusted odds ratio
[OR]=15.3, 95% confidence interval [CI]=4.4, 64.6) and not being told to go to the hospital
immediately (adjusted OR=5.3, 95% CI=1.9, 16.3). Receiving both these interventions reduced the
risk of not accessing hospital to 19%, from 96% for those who received neither intervention. Several
indicators of illness severity, including caregivers ranking of their childrens illness severity, the
presence of severe illness signs, and mortality, were investigated and found to not be important
explanatory factors.
Conclusion: Providing a referral slip and counseling the caregivers of severely ill children to go to the
hospital immediately appear to be powerful tools for increasing successful referral outcomes.
Key words: Integrated Management of Childhood Illness, IMCI, referral, referral constraints

The World Health Organization (WHO) and UNICEF in 1992 launched the Integrated
Management of Childhood Illness (IMCI) approach to providing health care and reducing the
mortality of young children. The IMCI provides case management guidelines that assist health
workers at first-level facilities to diagnose and manage the conditions thought to be
responsible for 70% of child deaths in less developed countries (1), including serious bacterial
infection, diarrhea, and low weight or feeding problems in young infants from 1 week up to 2
months of age; and pneumonia, diarrhea, malaria and severe febrile illnesses, ear problems,
malnutrition and anemia in older infants and children from 2 months up to 5 years old.
A key strategy of the IMCI approach is the identification and urgent referral to hospital of
severely ill infants and children. (Alternative guidelines are provided for areas where referral is
not possible.) Studies have found the guidelines to have good to moderate sensitivity for
detecting children who require hospital admission (2,3,4,5). However, the success of the IMCI
referral strategy depends on additional factors. Health workers must actually refer the
severely ill children identified by the guidelines. In addition to their level of clinical
competence, health workers referral behavior may be influenced by their assessment of
whether a childs caregiver is likely to follow their advice, and by the quality of communication
between first-level and referral facilities in the health system (6,7). It is also possible that
health workers may judge some referable children to be less severely ill than others and, in
such cases, reduce their counseling efforts or otherwise modify their referral practices. The
IMCI training curriculum attempts to overcome these problems by instructing health workers
in a standardized counseling method for the mothers of all referred children (8). This includes
explaining the need for referral, helping the mother to identify and overcome barriers to
adherence, and providing a referral note for the mother to take to the hospital.
Once a child is referred, the final step to a successful outcome is that the caregiver both
accepts the referral and is able to access the hospital. Recently documented referral
completion rates of from 24% to 48% (9,10,11) point to the difficulties of achieving this
objective. Furthermore, few studies have assessed which factors are most likely to constrain
adherence to referral advice for severely ill children, and thus should be emphasized by the
IMCI training. Constraints that have been identified include female gender of the sick child
(10), caregivers anxiety about the hospital (12,13), their perceived low severity of the illness,
their other childcare responsibilities, needing their husbands permission to make the journey,
associated costs and illiteracy (13), and health workers poor communication skills (12).
However, other than female gender, all these constraints were identified by purely descriptive
means.
In addition, 3 comparative studies identified barriers to seeking medical care for non-fatal
(14) and fatal child illnesses (14,15,16). However, careseeking constraints are likely to differ
for caregivers who have already entered the formal health care system, and careseeking
behavior may be altered by the act of referral itself. Health worker training in the use of case
management guidelines and health systems changes undertaken by districts using the IMCI
approach may also affect the referral advice given to caregivers. The present study sought to
identify modifiable factors that constrain adherence to referral advice, in a setting where the
IMCI approach has been fully implemented and hospitals are geographically accessible. The
levels of several key referral indicators in this setting were also assessed.

3DWLHQWVDQG0HWKRGV

Study Setting and Health Services

Imbabura Province is located in the plateaus and highlands of the Andes Mountains in
northern Ecuador. The 1998 under-5 years old population was reported as 40,807 (17) and
the corresponding official 1996 mortality rate, which is thought to be undercounted by about
20%, was 38.6/1000 births (18). The population is poor overall, but 88% of women with a
child under 2 years old are literate (19). Paved roads connect most villages to the major
towns, but high mountain passes, rain and mud slides can make the roads impassable at
times. Public buses run between the villages and towns, but are scheduled in remote areas
only once or twice per day at most. The Ministry of Public Health (MOH) operates 53 firstlevel facilities and 4 hospitals in the province. Most first-level facilities are centrally located
near the major towns, but there are also several sub-centers and health posts in the outlying
areas. The 4 hospitals are all centrally located.
The IMCI approach was implemented in the province in 1996-1997. All physicians, nurses
and nurse auxiliaries working in the MOH facilities were trained in the use of the guidelines.
To prepare for the present study and to help assure that children meeting referral criteria were
identified and referred, all health personnel were informed about the studys purpose and
underwent refresher training in the use of the IMCI guidelines. During the study, there was no
additional supervision of the health workers application of the guidelines, including their
counseling of caregivers to help identify and overcome referral constraints or their providing a
referral slip, beyond that normally accorded by the health care system.

Study Design
The study was conducted from September 1, 1999 to April 30, 2000 in 51 of the 53 MOH firstlevel facilities and in all 4 hospitals. All children from 1 week up to 5 years old seen at a firstlevel facility with an IMCI diagnosis were registered in a study log that included the type of
provider seen and the childs age, sex, diagnosis and referral status. Study logs were also
kept at the 4 hospital emergency departments, in which were recorded the children seen with
an IMCI diagnosis and each childs referral (referred by an MOH facility or not) and admission
status.
The interviewers were young, Ecuadorian women who had completed at least their
secondary education and had prior experience in administering structured interviews. They
tracked the children who were urgently referred to hospital for an IMCI diagnosis, with an
attempted follow-up period of 24 hours after the referral. The interviewers first searched the
hospitals, and if a referred child was not located the interviewer next went to the home. All
study subjects were read an approved informed consent statement and gave their consent to
participate in the study.
A structured questionnaire was used to ask caregivers about their childs illness, related
careseeking, and 6 categories of potential referral constraining factors: demographics and
socioeconomic status; family dynamics related to decision-making; indicators of illness
severity; the caregivers interaction with the health system, including her prior experience with
the referral hospital and any counseling provided by the first-level health worker about the
current illness; the caregivers perceived problems in accessing the hospital; and possible
physical and financial barriers, including geography, transportation and costs related to the
illness. Following the interview, caregivers of children who had not yet accessed the hospital
were offered assistance in reaching the hospital at that time.

The study underwent ethical review and was approved by the National Institute of
Scientific Research and Technological Development of the Ecuador Ministry of Public Health,
and by the Committee on Human Research of the Johns Hopkins University Bloomberg
School of Public Health.

Statistical Analyses
Referral indicators were calculated from the numerator and denominator data recorded in the
study logs. Referral indicators included the percent of children seen at a first level facility with
an IMCI diagnosis who were urgently referred, the percent of the referred children who
accessed hospital, and the percent of the referred children who accessed hospital who were
admitted.
Potential referral constraining factors were compared for the referred children who had
accessed and not accessed hospital by the time of the follow-up interview. Any variable with
an unadjusted odds ratio >1.5 or <0.67, or a chi-square or Fishers exact test p-value <0.20,
was allowed to enter logistic regression models built in a forward step-wise fashion. Factors in
the final model that had a multiplier effect on, or interacted with, each other were further
explored with 4 x 2 tables to evaluate the impact of their relationship on hospital access.
"Intervention bias" in the health workers application of their counseling efforts was
assessed by the chi-square or Kruskal-Wallis H-test to examine the distribution of indicators
of illness severity among children who received and did not receive any health worker actions
remaining in the final model. We also assessed clustering of these health worker actions in
order to determine whether particular first-level facilities might account for a greater proportion
of the high-impact interventions than would be expected by chance. Random effects logistic
regression models were fit to test the null hypothesis of zero intra-class correlation, with the
health facility as the independent variable and the health worker actions as the dependent
variables. Epi Info (20), SAS (21), LogXact (22) and Stata (23) software were used for data
entry and analysis.

5HVXOWV

Subjects and Referral Indicators

From September 1, 1999 to April 30, 2000, there were 11,668 visits to the 51 first-level study
facilities by children 1 week up to 5 years old with an IMCI diagnosis; 6181 (53%) of the visits
were by males, and 5487 (47%) were by females. One hundred seventy (1.5%) of the
children were urgently referred to hospital from 43 of the 51 first-level facilities. The referral
rate ranged from 0.0% to 6.2%. Eight referred children were lost to follow-up and 2 more died
before the interview could be conducted.
The caregivers of 160 referred children were interviewed; 67 (41.9%) of the children did
not access hospital, and their caregivers were interviewed with a median follow-up time after
the referral of 2 days (range, 0-9). The 93 referred children who accessed hospital did so
within a median time of 0 days (range, 0-3), and 67 (72%) of these children were admitted.
The caregivers of 53 admitted children were interviewed in the hospital with a median followup time of 1 day (range, 0-12). The caregivers of the other 14 admitted children plus the 26
who accessed hospital but were not admitted were interviewed outside of the hospital with a
median follow-up time of 2 days (range, 0-8).
A companion study of childhood deaths in the province during the same 8-month study
period determined that 11 (6.9%) of the 160 referred children died from their acute illness
after the interview had been completed. This included 7/93 (7.5%) and 4/67 (6.0%) of the
children who, respectively, did and did not access hospital (OR 1.28, P=0.76). The 2 children
who died before the referral interview could be conducted were captured by the mortality
study; 1 child had accessed a hospital and the other had not.
Table 1 shows some basic characteristics of the 160 referred children and their families.
The families were poor and the childrens caregivers had little formal education. Most
caregivers did not work outside of the home. Relative to expected disease severity by age
group, young infants were under-represented among the referred children. This reflected the
age distribution of the children seen at the first-level facilities with an IMCI diagnosis: 480
(4.1%) of the 11,668 visits were by infants from 7 days up to 2 months old. As shown by the
childs birthplace, almost 1/3 of the families had some prior experience with a hospital. The
median time it took the caregivers to reach a hospital by their preferred transportation method
suggests that physical access is not a major constraint in this setting.

Referral Constraints
Nineteen potential referral constraining factors, with at least 1 variable in each of the 6
categories of factors examined by this study, achieved a required cutoff level for inclusion in
the multivariate analysis (Table 2). All but 2 of these factors, fathers education less than 2
years and caregivers work obligations were positively associated with not accessing a
hospital. Some types of variables were asked about in multiple ways in order to assess both
the caregivers perception of the potential problem as well as to try to determine the factual
basis for a constraint. For example, caregivers were asked if the transportation cost to reach
the hospital was a problem for them, and they were asked the actual travel cost to the
hospital. The signs included in caregiver did not report a severe illness sign were based on
the IMCI guidelines, as well as signs that are alarming to caregivers in Ecuador. For example,
groaning (quejaba) is a sign of grave illness meaning that the child is too weak to emit strong
pain sounds. All the signs were spontaneously reported in response to an open-ended
question.

Variables that did not reach the threshold for entry into the multivariate analysis
included, among others, month of the year, childs gender, childs birthplace (home or
hospital), caregivers education, caregivers work outside the home, fathers occupation,
household crowding, caregivers other childcare responsibilities, seeking non formal health
care (friend, neighbor, pharmacy, traditional healer) before going to the first-level facility,
caregivers not ranking the illness as severe before seeing the first-level health worker,
illness duration, and all costs related to the illness other than those for transportation to the
hospital and for food and lodging while at the hospital. Some of these factors were common
problems but were similar for caregivers who accessed and did not access hospital. For
example, 70/93 (76%) and 51/67 (75%) caregivers who, respectively, did and did not access
hospital said that the hospital costs were a problem for them (OR 1.05, P=0.92).
Table 3 shows the logistic model, consisting of 6 risk factors, which was most highly
predictive of not accessing referral care. First-level health workers giving a referral slip and
advising caregivers to go to the hospital immediately had a strong multiplier effect on each
other, such that receiving both interventions versus receiving neither decreased a childs risk
of not accessing hospital from 96% to 19% (p<0.001) (Table 4). The last 2 terms of the
logistic model express that caregiver must stay overnight to access the hospital was a
significant referral constraint, and among those caregivers having to stay overnight, childs
age less than 2 months was associated with a greatly elevated risk. There was a synergistic
interaction between these 2 factors, such that a caregiver with both had an odds of not
accessing hospital 79 times as high as a caregiver who did not have to stay overnight.

Illness Severity
The relationship between the 2 health worker actions of providing a referral slip and
counseling to go to the hospital immediately, and the outcome of increased hospital access,
could be biased if the health workers preferentially performed these behaviors when seeing
the most severely ill children, whose caregivers might be more likely to seek hospital care
even in the absence of these interventions. Table 4 shows that the presence of severe illness
2
signs was comparable in children who received both or neither intervention (X 1df=0.72,
2
P=0.40), as was the risk of death (X 1df =0.03, P=0.87) and the interval between the first-level
visit and death (KW-H=0.00, P=1.00).
The multivariate analysis included several measures of illness severity. Caregiver did not
report a severe illness sign and caregiver did not rank her childs illness as severe after
seeing the first-level health worker were included because their unadjusted odds ratios were
>1.50 (Table 2). The type of health worker (physician vs. other) who saw the child did not
meet the required statistical threshold for inclusion in the logistic regression. Nevertheless, we
included it due to the concern that physicians might use their own diagnostic criteria beyond
the IMCI guidelines and refer only the most severely ill children. However, once first-level
health worker did not give a referral slip and first-level health worker did not say "Go to the
hospital immediately" came into the stepwise-constructed logistic model, all 3 severity
variables were prevented from entering the model.
The results from the random effects logistic regression models showed that there was no
significant clustering by health facility of these interventions. The p-values for first-level health
worker did not give a referral slip and first-level health worker did not say "Go to the hospital
immediately" were 0.36 and 0.11, respectively, even without adjustment for any other facilitylevel covariates such as patient mix.

High Risk Sub-Groups

We explored the impact of being given a referral slip and counseling to go to the hospital
immediately on the 2 high-risk sub-groups identified by the analysis. Receiving both these
interventions, versus receiving neither, decreased the risk of not accessing hospital from
100% to 16% (p<0.001) in children 2 months or older whose caregivers had to stay overnight
(Table 5). There were not enough young infants under 2 months old to definitively evaluate
the effect of these interventions in this highest risk group, but there was no dramatic
difference in access between those who received both versus neither intervention (Table 5:
29% vs. 0%, P=1.00). This was despite the fact that 4 caregivers of the 5 young infants who
received both interventions but did not access hospital, ranked their childs illness as severe
after seeing the first-level health worker. The other referral constraining factors identified by
the logistic model were common in the young infants whose caregivers had to stay overnight:
11/12 had some transportation cost and 7/12 mothers were not the decision-maker about
seeking referral care. However, the sample size of young infants was small, and these were
not significant constraints in this sub-group.

'LVFXVVLRQ



It is common in less developed countries for a large percentage, often the majority, of children
referred from first-level health facilities to not access hospital. This undermines the
effectiveness of health systems in general, and more specifically of the IMCI approach that is
being widely implemented in less developed countries. A significant part of IMCIs impact on
mortality will come from mothers accepting the referral and taking their children to a higher
level facility for care. The WHO IMCI training curriculum instructs health workers to give the
mothers of all referred children a referral note and to help them identify and overcome barriers
to accessing the hospital. However, data from Kenya show that of all the skills taught by IMCI
training, health workers performed least well in counseling (24).

Health Worker Actions

The present study found that giving a referral slip and advising caregivers to go to the hospital
immediately after departing the first-level facility greatly increased the likelihood that urgently
referred children would access hospital. These interventions multiplied each others effect to
increase successful referral outcomes. Prior researchers have documented the proportion of
referrals from first-level care to hospital that are written (6), but have not evaluated the impact
of giving a referral slip on caregiver behavior. Previous work has also shown that poor health
worker interpersonal communication skills can decrease the referral completion rate (9,12),
but the impact of specific messages, such as to go to the hospital immediately, has not been
examined. Other counseling efforts assessed by our study, such as helping caregivers to
overcome barriers to hospital entry and explaining the childs diagnosis to the mother, were
not as effective in ensuring a successful referral.

Illness Severity
Perception that an illness is not severe has been shown to delay seeking of formal medical
care (14) and the response to a referral directive (11). In our study, caregivers ranking of
illness severity was increased by the visit to the first-level facility, and was related to receiving
a referral slip and being counseled to go to the hospital immediately, but was independent of
caregivers response to the referral advice. Even caregivers who did not perceive their childs
illness to be severe were more likely to access the hospital if they were given a referral slip
and told to go immediately after leaving the first-level facility. Three additional severity
indicators, including health worker category, severe illness signs, and mortality, were
investigated and also found to not be important explanatory factors.

Young Infancy and Staying Overnight

Another referral constraint that might be amenable to counseling by health workers was
identified. Children whose caregivers had to spend the night away from home in order to
complete the referral were at increased risk of not being taken to hospital. Infants younger
than 2 months old whose caregivers had to stay overnight faced a particularly high risk. Being
given a referral slip and told to go to the hospital immediately was sufficient to overcame this
risk in the older infants and children, but these interventions did not modify the risk for the
young infants. De Zoysa et al previously found that mothers were reluctant to take their young
infants to hospital after referral, and felt that this was at least partly due to concern for their
special vulnerability (11). This descriptive study was conducted in a setting with an
unspecified but apparently unacceptable distance of available hospitals from the community,



so many of the caregivers would likely have had to stay overnight in order to complete the
referral. Their reluctance, and that of the caregivers in our study, could be due to concern for
their young infants frailty but might also indicate less willingness to expend additional
resources for their care.
The mothers in de Zoysa et als study mentioned transportation cost and needing their
husbands support and permission to make the journey as additional barriers to completing a
referral (11). However, Bhandari et al found a referral acceptance rate of only 24% for young
infants in the same setting, despite an offer of free transportation and hospital care (8). In our
study, most caregivers of young infants who had to spend the night away had a problem with
transportation cost and were not the decision-maker about seeking referral care, but the
sample size was too small to draw any firm conclusions about these possible constraints.
Further research is needed to identify appropriate counseling messages that might overcome
the referral constraints for young infants whose caregivers must stay overnight in order to
access the hospital.

Other Referral Constraints

The mother not being the decision-maker about hospital care, and transportation cost, were
significant referral constraints for our study group as a whole. De Zoysa et als study of young
infants supports these results, as does Terra de Souza et als finding that transportation costs
delayed the first seeking of formal medical care for fatally ill postneonates (14). These
constraints are unlikely to respond to health worker counseling. Health messages could be
developed to educate men about the importance of womens decision making for childrens
health care, and transportation cost might be attacked by a multi-sector intervention. Other
costs that we examined, including all the pre-hospital costs and the caregivers self-perceived
problems with hospital costs, were not associated with decreased access to hospital.
We did not find certain referral constraints that have previously been identified, including
female gender (10), anxiety about the hospital, other childcare duties and illiteracy (11). More
first-level (6,181/11,668 or 53%) and hospital (1,385/2,466 or 56%) visits in our study were for
males, but this could reflect differential illness rates. There were signs of anxiety about the
hospital in our study group. First-level health worker did not discuss how to gain entry to the
hospital and caregiver did not feel she could explain her childs illness to a hospital doctor
were risks for not completing the referral before adjustment for other variables. This anxiety
may often have been overcome by prior experience: 51 children were born in a hospital and,
of 37 caregivers who had taken their child to the local MOH hospital in the past 3 years, 20
(54%) were satisfied with the care they received. In the end, only 25/160 (16%) caregivers
said they were dissatisfied with the available hospital, and 9 of these completed the referral
anyway.
Other childcare responsibilities was also eliminated as a referral constraint by our use of
a comparative methodology: 45% of caregivers said this was a problem for them, but this was
split between the 39/93 (42%) who completed and 33/67 (49%) who did not complete the
referral (OR 1.34, P=0.45). There was also no difference between the number of children
cared for by those who did (median 2, range 06) and did not (median 2, range 07) complete
the referral (P=0.71). We did not directly measure literacy among our study group, but even
'caregiver's education less than 2 years' did not increase the risk for not completing the
referral. In de Zoysa et al's study in India, illiteracy had very practical consequences, such as
that mothers could not read the bus numbers (11). Literacy may be less important to referral
completion in the environment faced by mothers in Ecuador.



Study Limitations
Possible limitations of the study included a caregivers response bias. Caregivers of children
who did not reach hospital and whose condition had worsened at the time of the interview
might have tended to exaggerate the problems they faced in seeking hospital care. However,
only 8 children were reported to be feeling worse at the time of the interview and only 2 of
these had not accessed a hospital. Another possible bias was that health workers might have
modified their actions for mothers of referred children whom they judged to be less severely
ill. Health workers might also have concluded from their interaction with caregivers whether
they intended to complete the referral and, if this was seen as unlikely, withheld a referral slip
if these were in short supply. We were able to assess the potential bias related to judgement
of illness severity. The distributions of severe illness signs and mortality among the children
who did and did not receive a referral slip and counseling to go to the hospital immediately
suggest that the health workers did not treat referred children differently based on their illness
severity. A final limitation is that the study findings may not be fully generalizable to other
settings where, for example, the road system is less developed and geographic access may
therefore be more difficult.

Conclusions
42% of the referred children in this study were not taken to hospital, despite the fact that as a
group they were very sick: 6.9% died from their illness, and 72% of those who accessed a
hospital were admitted. Six factors were associated with not completing the referral, and the
findings suggest that health worker training in referral counseling should be strengthened.
Health systems should ensure that referral slips are available, and emphasize to health
workers that they should give a slip and counsel the caregiver of each urgently referred child
that the severity of the illness requires going to the hospital immediately after leaving the firstlevel facility. Health workers should be made aware that young infants whose caregivers must
spend the night away from home in order to complete the referral might be at particularly high
risk of not reaching the hospital. They should be trained to assess and address this
constraint. Other referral constraining factors identified by the study, such as transportation
cost and the mother not being the primary decision-maker regarding hospital care, may be
less amenable to intervention by the health system.

$FNQRZOHGJPHQWV



This work was supported by the US Agency for International Development through the Johns
Hopkins Family Health and Child Survival Cooperative Agreement and through the BASICS
Project.

5HIHUHQFHV



WHO/UNICEF. The WHO/UNICEF Approach to Integrated Management of the Sick Child.

Geneva: WHO; 1995
Kalter HD, Schillinger JA, Hossain M, Burnham G, Saha S, deWit V, et al. Identifying Sick
Children Requiring Referral to Hospital in a Developing Country. Bull World Health Organ.
1997;75 Suppl 1:65-75
Weber MW, Mulholland EK, Jaffar S, Troedsson H, Gove S, Greenwood BM. Evaluation of
an Algorithm for the Integrated Management of Childhood Illness in an Area with Seasonal
Malaria in the Gambia. Bull World Health Organ. 1997;75 Suppl 1:25-32
Perkins BA, Zucker JR, Otieno J, Jafari HS, Paxton L, Redd SC, et al. Evaluation of an
Algorithm for Integrated Management of Childhood Illness in an Area of Kenya with High
Malaria Transmission. Bull World Health Organ. 1997;75 Suppl 1:33-42
Simoes EA, Desta T, Tessema T, Gerbresellassie T, Dagnew M, Gove S. Performance of
Health Workers After Training in Integrated Management of Childhood Illness in Gondar,
Ethiopia. Bull World Health Organ. 1997;75 Suppl 1:43-53
Nordberg E, Holmberg S, Kiugu S. Exploring the Interface Between First and Second Level
of Care: Referrals in Rural Africa. Trop Med Int Health. 1996;1:107-111
DiPrete L, Valadez J. Assessing growth monitoring services in the Costa Rican Primary
Health Care System. PRICOR Country Report Series: Costa Rica Systems Analysis of
Growth Monitoring. 1991. Aug 15
WHO/UNICEF. IMCI Training Curriculum: Counsel the Mother, Section 5: Refer the Child.
Geneva: WHO; 1996
Bhandari N, Bahl R, Bhatnagar V, Bhan MK. Treating Sick Young Infants in Urban Slum
Setting [letter]. Lancet. 1996 Jun 22;347(9017):1774-1175
Bakry N, Laabid A, De Brouwere V, Dujardin B. Why Patients Do Not Comply with Reference
Decision Made by General Practitioners? Rev Epidemiol Sante Publique. 1999. Oct;47
Suppl 2:2S65-74
Ganatra B, Hirve S. Male Bias in Health Care Utilization for Under-fives in a Rural
Community in Western India. Bull World Health Organ. 1994;72:101-104
Rowe AK, Onikpo F, Lama M, Cokou F, Deming MS. Management of Childhood Illness at
Health Facilities in Benin: Problems and Their Causes. Am J Public Health.
2001;91:1625-35
de Zoysa I, Bhandari N, Akhtari N, Bhan MK. Careseeking for Illness in Young Infants in an
Urban Slum in India. Soc Sci Med. 1998 Dec;47(12):2101-11
Van der Stuyft P, Sorensen SC, Delgado E, Bocaletti E. Health Seeking Behavior for Child
Illness in Rural Guatemala. Trop Med Int Health. 1996;1:161-170
Sutrisna B, Reingold A, Kresno S, Harrison G, Utomo B. Care-seeking for Fatal Illnesses in
Young Children in Indramayu, West Java, Indonesia. Lancet. 1993;342(8874):787-9
Terra de Souza AC, Peterson KE, Andrade FM, Gardner J, Ascherio A. Circumstances of
Post-Neonatal Deaths in Ceara, Northeast Brazil: Mothers' Health Care-seeking Behaviors
During their Infants' Fatal Illness. Soc Sci Med. 2000 Dec;51(11):1675-93
Divisin Nacional de Estadsticas. Ministerio de Salud Pblica: Quito, Ecuador; 2000.
Anuario de Estadsticas Vitales, Nacimientos y Defunciones. Ministerio de Salud Pblica:
Quito, Ecuador; 1996.



BASICS. Encuesta de Hogares para AIEPI. Provincia de Imbabura, Ecuador. BASICS:

Arlington, VA; 1999 April. p21.
Dean AG, Dean JA, Coulombier D, Brendel KA, Smith DC, Burton AH, et al. Epi Info, Version
6: a word-processing, database, and statistics program for public health on IBMcompatible microcomputers. Atlanta, GA: Centers for Disease Control and Prevention;
1995.
The SAS Institute. The SAS System, Version 8.0. Cary, NC: The SAS Institute; 1999.
CYTEL Software Corporation. LogXact Version 4.1. Cambridge, MA: CYTEL Software
Corporation; 2000.
Stata Corporation. Stata Version 6.0. College Station, TX: Stata Corporation; 1999.
Odhacha A, Orone H, Pambala M, Odongo R, Lungapher J, Kituye A, et al. Health Worker
Performance After Training in Integrated Management of Childhood Illness--Western
Province, Kenya, 19961997. MMWR Morb Mortal Wkly Rep. 1998. Nov 27;47(46):9981001.



Table 1. Characteristics of 160 referred children and their families,

Imbabura Province, Ecuador, September 1, 1999April 30, 2000
Characteristic

Childs age

7 days -1 month
2-11 months
1-4 years

Childs gender

Male
Female

Childs birthplace

Home
Hospital
Other
Unknown

Childs caregiver
Education
Work outside the home
Mother

(%)

25
65
70

(15.6)
(40.6)
(43.8)

93
67

(58.1)
(41.9)

102
51
6
1

(63.8)
(31.9)
(3.8)
(0.6)

160

Household
Water supply
Floor material
Persons/sleeping room

No indoor water
Earthen
median (range)

Time to reach a hospital

median minutes (range)

(100.0)
4 (0-14)
0 (0-63)

median years (range)

median hours/week (range)
149

(93.1)

123
82

(76.9)
(51.3)
4 (1.7-11.0)
30 (3-360)



Table 2. Potential factors constraining adherence to referral advice, with

an unadjusted odds ratio >1.5 or <0.67, or a p-value <0.20

Potential referral constraining factor

Accessed
hospital
N (%)

Did not
access
hospital
N (%)

Odds
ratio

Demographics
Fathers education <2 years
Neighborhood residence <5 years
Residence in Ibarra Cantn

21 (23.6)
44 (47.3)
17 (18.2)

5 (8.6)
39 (58.2)
17 (25.4)

0.31
1.55
1.52

0.03
0.20
0.33

Family dynamics
Mother not the decision-maker about seeking referral care

34 (36.6)

41 (61.2)

2.74

<0.01

Caregiver's perceived problems

Child not sick enough to go to hospital
Not satisfied with available hospital
Work obligations
Cost of food and lodging while child at hospital

6 (6.5)
9 (9.7)
16 (17.2)
46 (49.5)

13 (19.4)
16 (23.9)
5 (7.4)
43 (64.2)

3.49
2.93
0.39
1.83

0.02
0.03
0.10
0.08

Hospital access
Transportation to hospital was via walking or bus
Caregiver must stay away overnight to access the hospital
Transportation cost to access hospital >0 sucres
Time >1 hour to reach hospital

72 (90.0)
58 (62.4)
68 (73.1)
23 (27.1)

65 (97.0)
59 (90.8)
62 (92.5)
24 (42.9)

8.13
5.93
4.56
2.02

<0.01
<0.01
<0.01
0.07

59 (63.4)

50 (74.6)

1.69

0.16

8 (8.7)
21 (22.6)

36 (54.5)
42 (62.7)

12.60
5.76

<0.01
<0.01

20 (21.7)

34 (53.1)

4.08

<0.01

27 (29.0)

36 (53.7)

2.84

<0.01

13 (14.3)

16 (25.4)

2.04

0.10

17 (18.3)

18 (26.9)

1.64

0.25

Illness severity
Caregiver did not report a severe illness sign (very sleepy,
vomiting everything + age >2 months, drinking
poorly/not able to drink or breastfeed, groaning, fast
breathing + age <2 months, fever + age <2 months,
subcostal indrawing, draining ear pus + age <2 months,
convulsions)
Health system-caregiver interaction
First-level health worker did not give a referral slip
First-level health worker did not say "Go to the hospital
immediately"
First-level health worker did not discuss how to gain entry to
the hospital
Caregiver did not rank her child's illness as 'severe' after
seeing the first-level health worker
Caregiver did not feel she could explain her child's illness to
a hospital doctor after seeing the first-level health
worker
Caregiver did not know her child's diagnosis after seeing
the first-level health worker



Table 3Logistic regression model, showing the adjusted odds ratios, 95% confidence intervals
and exact p-values of the factors most highly predictive of not accessing hospital
Referral constraining factors
First-level health worker did not give a referral slip

Odds ratio
15.3

95% CI
4.4 - 64.6

Exact P
<0.001

Transportation cost to access hospital >0 sucres

6.4

1.4 - 38.6

0.012

First-level health worker did not say "Go to the

hospital immediately"

5.3

1.9 16.3

0.001

Mother not the decision-maker about seeking

referral care

5.0

1.7 - 15.7

0.001

Caregiver must stay overnight to access the

hospital, and childs age <2 months

79.2

7.4 1,429.5

<0.001

Caregiver must stay overnight to access the

hospital, and childs age >2 months

7.0

1.8 34.0

0.002

9
(56.2)
27
(96.4)

7
(43.8)
1
(3.6)

First-level health First-level health worker said "Go to the

worker did not
hospital immediately"
give a referral
slip
First-level health worker did not say "Go to
the hospital immediately"

15
(44.1)

15
(18.8)

19
(55.9)

65
(81.3)

8
(28.6)

6
(37.5)

6
(17.6)

30
(37.5)

N
(%)

N
(%)

N
(%)

First-level health worker did not say "Go to

the hospital immediately"

First-level health First-level health worker said "Go to the

worker gave a
hospital immediately"
Referral slip

Severe
symptoms

Did not
access
hospital

Accessed
hospital

2
(7.1)

0
(0.0)

4
(11.8)

5
(6.3)

N
( %)

5.5
(38)

12
(229)

5
(115)

Median
(range)

Illness duration
after the visit

Deaths

Table 4Multiplier effect on adherence to referral advice of health workers giving cargivers a referral slip and counseling
them to go to the hospital immediately (not adjusted for other referral constraining factors); and the distributions of severe
symptoms and deaths among the referred children



Caregiver
must stay
overnight to
access
hospital, and
Childs age
>2 months

Caregiver
must stay
overnight to
access
hospital, and
Childs age
<2 months

2
(28.6)
0
(0.0)
0
(0.0)
0
(0.0)
41
(83.7)
10
(47.6)
5
(41.7)
0
(0.0)

First-level health worker said "Go to the hospital immediately"

and gave a referral slip
First-level health worker did not say "Go to the hospital
immediately" and gave a referral slip
First-level health worker said "Go to the hospital immediately"
and did not give a referral slip
First-level health worker did not say "Go to the hospital
immediately" and did not give a referral slip
First-level health worker said "Go to the hospital immediately"
and gave a referral slip
First-level health worker did not say "Go to the hospital
immediately" and gave a referral slip
First-level health worker said "Go to the hospital immediately"
and did not give a referral slip
First-level health worker did not say "Go to the hospital
immediately" and did not give a referral slip

N
(%)

Accessed
hospital

22
(100.0)

7
(58.3)

11
(52.4)

8
(16.3)

2
(100.0)

2
(100.0)

1
(100.0)

5
(71.4)

Did not
access
hospital
N
(%)

Table 5. Impact on adherence to referral advice of first-level health workers

giving caregivers a referral slip and counseling them to go to the hospital
immediately, in children <2 months old and >2 months old whose caregivers
must stay overnight (not adjusted for other referral constraining factors)