Summary
Lancet 2008; 372: 910–16 Background Countries with high rates of child mortality tend to have shortages of qualified health workers. Little
See Editorial page 863 rigorous evidence has been done to assess how much the quality of care varies between types of health workers. We
See Comment page 870 compared the performance of different categories of health workers who are trained in Integrated Management of
*Members listed at end of paper Childhood Illness (IMCI).
Universidad Nacional Mayor de
San Marcos, Universidad Methods We analysed data obtained from first-level health facility surveys in Bangladesh (2003), Brazil (2000), Uganda
Peruana Cayetano Heredia and (2002), and Tanzania (2000). We compared the clinical performance of health workers with longer duration of
Instituto de Salud del Niño,
Lima, Peru (Prof L Huicho MD);
preservice training (those with >4 years of post-secondary education in Brazil or >3 years in the other three countries)
Department of Child and and shorter duration (all other health workers providing clinical care). We calculated quality of care with indicators of
Adolescent Health and assessment, classification, and management of sick children according to IMCI guidelines. Every child was examined
Development twice, by the IMCI-trained health worker being assessed and by a gold-standard supervisor.
(R W Scherpbier MD) and
Human Resources for Health
Department Findings 272 children were included in Bangladesh, 147 in Brazil, 231 in Tanzania, and 612 in Uganda. The proportions
(A M Nkowane MA), WHO, of children correctly managed by health workers with longer duration of preservice training in Brazil were 57·8%
Geneva, Switzerland; and (n=43) versus 83·7% (n=61) for those with shorter duration of training (p=0·008), and 23·1% (n=47) versus 32·6%
Universidade Federal de
Pelotas, Pelotas, Brazil
(n=134) (p=0·03) in Uganda. In Tanzania, those with longer duration of training did better than did those with shorter
(Prof C G Victora MD) duration in integrated assessment of sick children (mean index of integrated assessment 0·94 [SD 0·15] vs 0·88
Correspondence to: [0·13]; p=0·004). In Bangladesh, both categories of health worker did much the same in all clinical tasks. We recorded
Dr Luis Huicho, Universidad no significant difference in clinical performance in all the other clinical tasks in the four countries.
Nacional Mayor de San Marcos,
Universidad Peruana Cayetano
Heredia and Instituto de Salud
Interpretation IMCI training is associated with much the same quality of child care across different health worker
del Niño, Av Brasil 600, LI05 categories, irrespective of the duration and level of preservice training. Strategies for scaling up IMCI and other
Lima, Peru child-survival interventions might rely on health workers with shorter duration of preservice training being deployed
lhuicho@gmail.com in underserved areas.
Funding Bill & Melinda Gates Foundation and the US Agency for International Development.
Results
Adjusted mean difference (95%CI) p value
between health workers with shorter
Table 1 shows the different categories of health workers
and longer preservice training providing care in every country, the duration of their
Bangladesh* –0·01 (–0·05 to 0·02) 0·46
preservice training, and the number of children that they
Brazil† 0·19 (0·08 to 0·30) 0·001
saw. In Bangladesh, children were seen by medical
Tanzania‡ –0·06 (–0·09 to –0·02) 0·004
assistants and subassistant community medical officers,
Uganda§ 0·01 (–0·03 to 0·06) 0·51
both of whom received 4 years of post-secondary training,
and by family welfare visitors, who received 18 months of
Adjusted mean difference through general linear model. *Adjusted for child age, post-secondary training. Physicians did not manage sick
and child sex. †Adjusted for child age, child sex, municipality, and facility type.
children in the first-level facilities who were enrolled in
‡Adjusted for child age, child sex, district, and facility type. §Adjusted for child
age, child sex, and district category. the Bangladesh study site. In Brazil, IMCI was provided
by doctors and nurses who received 6 and 4 years of
Table 3: Index of integrated assessment of children by IMCI-trained
university training, respectively. In Tanzania, children
health workers
were managed by clinical officers and assistant clinical
officers (who received 2–4 years of post-secondary
Crude odds p value Adjusted odds p value training), as well as by maternal and child health aides
ratios (95% CI) ratios (95% CI)
(2 years of post-primary school training) and nurse
Correct classification assistants (1 year of post-primary school training). In
Bangladesh 1·28 (0·71–2·31) 0·41 1·28 (0·70–2·31) 0·42 Uganda, clinical officers (3 years of post-secondary
Brazil 0·58 (0·29–1·19) 0·14 0·39 (0·09–1·66) 0·20 training), nurses and midwives (18–36 months of
Tanzania 0·79 (0·38–1·64) 0·53 0·98 (0·39–1·82) 0·98 post-secondary training), and nurse assistants (3 months
Uganda 1·26 (0·85–1·86) 0·25 1·28 (0·87–1·90) 0·22 of post-primary training) accounted for almost all IMCI
Correct management care provision, which was complemented by physicians
Bangladesh 0·8 (0·45–1·45) 0·47 0·81 (0·45–1·47) 0·49 (5 years of university training) and other IMCI-trained
Brazil 0·27 (0·09–0·73) 0·008 0·11 (0·02–0·73) 0·02 health workers.
Tanzania 1·02 (0·86–1·21) 0·82 1·56 (0·70–3·49) 0·28 Table 2 shows the clinical performance of IMCI-trained
Uganda 0·62 (0·41–0·95) 0·03 0·59 (0·38–0·90) 0·01 health workers by type, grouped by duration and level of
preservice training. In Brazil, nurses did better than
Adjustment for confounding factors was made through logistic regression.
doctors did in the assessment and management of sick
Table 4: Odds ratios for correct classification and management by IMCI- children, whereas they both had much the same
trained health workers with longer duration of training compared with performance in the correct classification indicator
those with shorter duration of training
(table 2). In Uganda, health workers with shorter duration
of training did better in correct management than did
training) was entered in the model as the random factor, those with longer duration of training, but both categories
district or municipality as a fixed factor, and the covariates showed levels of performance that were well below those
were type of health facility, age of child, and sex of child. reported from the other countries (table 2). In Bangladesh,
We used logistic regression for dichotomous outcomes health workers with shorter duration of training did as
(correct classification and management). In Bangladesh,
information about type of health facility was not available. 100 Rural physicians
In Uganda, most information about facility type was Rural nurses
missing and could not be included in the adjusted model. 90
10
Role of the funding source
0
The sponsor of the study had no role in the study design, Brazil (2000) Honduras (2000) Algeria (2002) The Gambia Tanzania (2002) Uganda (2000)
data collection, data analysis, data interpretation, or (2003)
writing of the report. The corresponding author had full Country (latest year with data)
access to all the data in the study and had final Figure 1: Proportion of total physicians and nurses working in rural areas
responsibility for the decision to submit for publication. The red bars indicate the proportion of the population in rural areas in 2000. Data from reference 29.
In Brazil, nurses did better than did doctors in correct Table 5 shows the ratio of physicians and nurses, and
management, including counselling tasks on how to ranking of child deaths in the four study countries.
administer the drug at home. Doctors might leave this In some countries, existing regulations preclude health
task to clerks who dispense drugs in the facilities, thus workers with shorter duration of training from under-
failing to meet one of the criteria for correct taking tasks that they might appropriately do.25 For
management. Additionally, the Brazilian nurses also example, in Brazil, pressure from medical societies led
did better than their physician colleagues did in correct to discontinuation of IMCI training for nurses in 2002.
assessment. The Brazilian health system is much more In view of the health workforce crisis affecting
sophisticated than that of Bangladesh, Tanzania, or underserved areas, these are regrettable measures that
Uganda—Brazilian nurses receive equal or longer are further challenged by results from our study.
duration of training than do most health workers who However, in other countries—eg, Ghana, Malawi,
were classified in the group with longer training in the Tanzania, and Zambia—nurses were trained and allowed
other three countries. That we recorded no systematic to undertake tasks that were usually restricted to
advantages of longer training, despite these differences physicians, thus providing support for task shifting
in the composition of categories that we investi- practices long before the term was accepted.30 These
gated, strengthens the external validity of our overall analyses drew attention to the importance of proper
conclusions. regulation, supervision, and motivation to allow task
One of the limitations of the study is its observational shifting to succeed.30
design. However, randomly allocating trainees into Achievement of high and effective population
becoming nurses or doctors, for example, is challenging coverage is a key component in the efforts to reach a
both ethically and practically. The finding that the measurable effect with child-health interventions.31 As
performance of health workers with shorter training was argued by Friedman32 in the context of caesarean
similar to that of those with longer training is in contrast sections, a substantial increase in coverage with health
with this potential selection bias. Furthermore, workers with shorter training—even if they are less
characteristics of workers that might affect the clinical skilled in operative deliveries than are those with longer
performance—such as tenure and age of worker—could duration of training—can contribute to an important
be important, but data for these variables were not reduction in maternal mortality. A similar argument
available. can be made for the reduction of deaths in children
Our results need to be interpreted within a broader younger than 5 years, in view of the published work
conceptual framework of factors affecting clinical showing that case management of specific infections by
performance, including the availability of equipment and community health workers with short training is safe
supplies, supportive supervision, workforce motivation, and effective.33,34
economic incentives, and a positive personal and Our results support the principle of task shifting as an
professional environment.3,24 In the sites that were important component of the strategy for strengthening
included in the present analyses, equipment and drug health systems in areas that are understaffed. Training
availability were mostly adequate, but supervision tended and deployment of motivated health workers with shorter
to be irregular.9 However, we did not gather information duration of preservice training in regions with high
about motivation, incentives, or quality of the professional mortality in children younger than 5 years will help
environment. ensure that all children who need clinical care are seen
Although the ratio of physicians and paramedical by a qualified provider.
workers varies enormously from country to country,5,25 Contributors
nurses and other health workers tend to consistently LH, RWS, and CGV participated in the article conception, writing, and
outnumber doctors.26 Within countries, ratios are discussion. LH analysed the data. AMN took part in the writing and
discussion of the article. All authors saw and approved the final version
generally highest in urban areas, particularly for of the report.
physicians.27 In rural areas, where health-care needs are
The Multi-Country Evaluation of IMCI Study Group
greatest, the numbers of health workers, particularly Members who contributed to this paper, particularly in constructing
doctors, tend to be much less than population needs in table 1, include: João Amaral (Department of Maternal and Child Health,
most countries (figure 1).28,29 At the same time, the nurse Federal University of Ceará, Fortaleza, Brazil), Shams El Arifeen
to physician ratio in rural areas varies between 3 and 19, (International Centre for Diarrhoeal Disease Research, Dhaka,
Bangladesh), Joanna Armstrong Schellenberg (Ifakara Health Research
indicating a greater availability of nurses in rural areas and Development Centre, Ifakara, Tanzania and Gates Malaria
than in urban areas, even in countries where less than Partnership, London School of Hygiene and Tropical Medicine, London,
half of their physician and nurse workforce reside in UK), Eleanor Gouws (Epidemiology and Analysis
rural areas (figure 2). The differential distribution of Division, UNAIDS, Geneva, Switzerland), Theopista John (Department
of Family and Reproductive Health, WHO, Dar es Salam, Tanzania), and
doctors and nurses in figure 2 indicates that in George Pariyo (Department of Health Policy Planning and Management,
underserved, rural areas, health workers with longer Makerere University Institute of Public Health, Kampala, Uganda).
preservice training (such as doctors) are less available Pedro C Hallal from the Universidade Federal de Pelotas, Pelotas, Brazil
than are those with shorter training (such as nurses). provided assistance with the clustering analyses.