Introduction

The rational approach to health promotionthat information given by health workers during clinic based or community based contacts will bring about a change in health behaviouris still an integral part of primary healthcare strategies. ^{1 2} In practice, opportunities for one to one health education are given low priority by busy health workers. A survey of perinatal services across India reported that opportunities to give health education messages to mothers in the community were invariably missed.³

The effectiveness of health education has also been questioned. A recent review of over 500 articles about health education in developing countries found that only 11% described and evaluated actual attempts at health education. Of these, four described randomised studies and only three fulfilled the author's criteria for a rigorously designed evaluation.⁴ In countries with few resources there is also a trade off between impact and sustainability. Interventions that are considered successful usually result from small scale, well resourced projects which cannot be reproduced on a large scale. One non-randomised evaluation of an initiative to encourage postnatal health education in a district hospital in Bihar, India, did show significant improvements in early breast feeding practices, although health education by the health workers was not maintained in the longer term.⁵

With increasing use of hospital maternity and immunisation services, especially in urban areas of the developing world,⁶ perinatal contact with mothers represents an opportunity for health education about infant care and family planning. In developing countries 50-60% of infant deaths occur in the neonatal period,⁷ and mortality from acute respiratory infections is highest in the first two months of life, when a mother's response to warning signs is crucial for survival. Failure to use postnatal contraception may also lead to an early repeat pregnancy, with attendant risks to maternal health.

In our study, the prospectively defined hypothesis was that one to one postnatal health education for mothers would positively affect their subsequent knowledge of and practices about infant care and family planning. Because an intervention should be feasible and sustainable on a large scale, education was restricted to a maximum of two contacts. Clinical objectives were to evaluate the impact of the intervention on uptake of immunisation, knowledge about and care of acute respiratory infections and diarrhoea in infants, the duration of exclusive breast feeding, infant growth, and use of postnatal family planning services.

	Subjects and methods

The study was conducted from November 1994 to May 1996. Oral consent from mothers for inclusion in the study was obtained before assignment. The study received ethical approval from the Nepal Health Research Council.

Setting
Nepal is one of the least developed countries in the world, with an infant mortality of 98/1000 live births, maternal mortality of 1500/100 000 live births, 26% adult literacy, and a prevalence of contraceptive use of 23%.⁸ The estimated population of Kathmandu municipality is 500 000, with an annual urban growth rate of 7.4%.⁹ Prasuti Griha is the main government funded maternity hospital in Kathmandu, with 250 beds, 15 000 deliveries annually, and outpatient services for the local urban and surrounding populations. As there are no formal addresses in Kathmandu, a house to house survey of two communities was conducted before the study. Kirtipur is a periurban area 5 km south west of the hospital that contains 3663 households with a total population of 21 368. It is a settled community of mainly wage labourers and farmers. Kalimati is an urban area of central Kathmandu situated 2 km from Prasuti Griha and containing 2467 households with a total population of 13 875. This is a mixed community of long term residents and recent migrants.

View larger version (36K): [in this window] [in a new window]   Flow chart showing progress of mothers through randomised trial

Eligibility
All pregnant women admitted to Prasuti Griha hospital for delivery residing in these two communities were eligible for entry to the trial. Two mothers entered into the trial whose deliveries resulted in a stillbirth were withdrawn from the trial and received neither the intervention nor follow up.

Protocol
The health education intervention was developed with hospital staff in collaboration with consultants experienced in health education and women's development. Three female health educators, two midwives, and one community health worker were trained to give the health education. All were fluent in the two local languages, Nepali and Newari, and conducted the education intervention in the appropriate language. The health education session lasted about 20 minutes and was designed to be interactive and supportive rather than prescriptive in style. It was tested beforehand with 20 mothers, and modifications were made in the light of this experience. The health educators were monitored weekly during the trial by two principal investigators to check the quality of the intervention with regards to the content and the style of delivery, especially the level of interaction, and constructive critical feedback was given.

Outcome measures
After birth, data were collected on the pregnancy; mode and outcome of delivery; and infant gestational age, birth weight, length, and head circumference. Infant weight was measured to the nearest 50 g with a Soenhle electric infant weighing scale. Infant length was measured to the nearest 0.5 cm with a Rollametre (Child Growth Foundation), and head circumference was measured to the nearest 0.1 cm with a tape measure. Gestational age was assessed by the Parkin method.¹⁰

Sample size
To assess the baseline situation we reviewed recent national survey data and conducted a pilot survey at the hospital outpatient clinic of 200 postnatal mothers100 at birth and 100 at up to 4 months post partum.

Infant nutritional status If education helps to prolong the duration of exclusive breast feeding, nutritional outcome might be improved. Assuming infants of mothers receiving the intervention grew on average along the 50th centile for British infants, a difference of 300 g in weight at 6 months (3.8%) between the group receiving no education with the two groups receiving health education at birth would be detected with 95% confidence limits and a power of 80% with sample sizes of 131 and 262. 

Mothers' knowledge of managing infant diarrhoea and acute respiratory infection A Nepal national survey in 1994 found that only 37% of urban children with diarrhoea received oral rehydration solution.¹² In Pakistan few mothers spontaneously mentioned rapid breathing as a sign of pneumonia,¹³ and in the Philippines only 22% of cases of severe acute respiratory infection were recognised as severe by the mothers.¹⁴ We hypothesised that 40% of control mothers and 60% of intervention mothers would correctly describe the signs of pneumonia and how to manage diarrhoea, requiring 107 mothers in each group (95% confidence interval, power 80%).

Immunisation uptake In 1991 in Nepal 74% of children aged 12 months had received three doses of diphtheria and pertussis vaccine and oral polio vaccine, and 81% were vaccinated against tuberculosis.¹⁵ We hypothesised that 40% of control and 60% of intervention infants would be fully immunised by 6 months of age, requiring 107 mothers in each group (95% confidence interval, power 80%).

Family planning In our pilot survey 20% of postnatal mothers were using a method of contraception at 4 months. Use of contraceptives by currently married women nationally was estimated at 14%,¹⁵ but this can be assumed to be higher in urban areas. We hypothesised a 20% uptake of contraception in the control group by 6 months post partum and an improvement to 33% in the intervention group, requiring 195 mothers in each group (95% confidence interval, power 80%).

Using these figures, we enrolled 540 subjects in order to compare four subgroups: mothers receiving health education immediately after birth and at 3 months post partum (group A), health education at birth only (group B), health education at three months only (group C), or no health education at all (group D). For outcomes at three months, we combined groups A and B as the intervention group and C and D as the control group. For the outcomes at 6 months, the groups were compared individually.

Randomisation and blinding
The unit of randomisation was the individual mother. Restricted randomisation was used in blocks of 20, each block consisting of a random ordering of the numbers 0-19. Numbers 0-4, 5-9, 10-14, and 15-19 were assigned to groups A to D respectively. The details of allocation to groups for consecutively recruited mothers were in sealed envelopes. Timing of assignment was when a mother was identified by the research team either in labour or shortly after delivery. A member of the research team checked the hospital admission register at least twice each day between 7 am and 8 pm. The generator of the assignment was not involved in the execution of the allocation. There were no prospectively defined rules for stopping the trial.

Statistical analysis
To estimate the effect of the trial intervention between the groups we measured the mean differences, 95% confidence intervals, and P values for continuous data, and the odds ratios, 95% confidence intervals, and P values for categorical data. We used the Mantel-Haenszel test to check for heterogeneity of categorical data, giving ² and P values, and analysis of variance for continuous data, giving F values and P values. We analysed data on an intention to treat basis in which we compared intervention and control groups irrespective of the quality of the education intervention. For statistical analysis, we used computer software Statview version 4.0 and Stata version 5.0. 

	Results

Subjects
The figure shows the details of participant flow and follow up. We recruited 540 mothers, 135 to each of the four groups, and followed up 403 (75%) to 3 months post partum and 393 (73%) to 6 months. The main reason for loss to follow up was the mother moving back to her parental home as part of cultural tradition. Table 1 shows the baseline characteristics of the mothers and infants.

Outcome at 3 months
Table 2 shows the outcomes at 3 months post partum. We compared mothers in groups A and B, who received health education at birth, with those in groups C and D, who received none. Mothers in groups A and B were slightly more likely to report tachypnoea as a sign of acute respiratory infection, but this did not quite reach statistical significance (odds ratio 1.48, 95% confidence interval 1.00 to 2.19, P=0.06). Also, 20% of mothers in groups A and B were using contraception compared with only 14% of those in groups C and D, but this difference was not significant. There were no differences for the other outcomes.

Outcome at 6 months
Table 3 shows the outcomes at 6 months post partum. We made two broad comparisons: groups A and B (health education at birth) compared with groups C and D (no health education at birth), and groups A and C (health education at 3 months) compared with groups B and D (no health education at 3 months). The only significant difference we observed for all outcomes was an increase in uptake of family planning at 6 months in groups A and B (odds ratio 1.62, 95% confidence interval 1.06 to 2.5). To test for interactions, we compared outcomes by health education at birth stratified by whether health education was given at 3 months post partum using tests for heterogeneity: we found no significant interactions.

	Discussion

This trial in Nepal has shown that a health education intervention (one to one counselling of mothers by health educators) given on two occasions, immediately after delivery and 3 months later, had no significant impact on the mothers' knowledge and practices of child care or infant health outcomes, but there was a slight improvement in uptake of family planning at 6 months after birth. Given the higher than expected level of immunisation in all groups, we cannot rule out the possibility that health education may have had an impact in situations where coverage is lower.

Trial design
Our study included only women who chose institutional delivery. Whether mothers who gave birth at home would benefit from health education more than those who gave birth at hospital is questionable, but it is difficult to target mothers delivering at home and to conduct a trial of intervention in the home.

Evaluation of health education interventions
Recommendations for the design of health education interventions and the importance of including evaluation in health education programmes have been widely reported. ^{4 17-20} For example, the American Public Health Association stated that "from the outset, a health promotion program should be organised, planned and implemented in such a way that its operation and effects can be evaluated."¹⁷ In practice, however, evaluation is rare. In a review of health education in developing countries spanning 10 years, only 11% of published articles described and evaluated the health education programme.⁴ Most of the evaluations were methodologically unsound so firm conclusions could not be drawn about the overall efficacy of health education. Randomised controlled studies, the ideal design, have rarely been reported from developing countries: only four of the studies reviewed by Loevinsohn used a randomised controlled design, and one of these failed to meet other of his criteria for an adequate study design.⁴

Conclusions
Our results indicate the need for further, well designed evaluations of health education interventions that are randomised and controlled, provide a clear definition of aims, and present pre-intervention and post-intervention data for carefully defined outcome measures. Future evaluations of education interventions also need to explore, through qualitative research, the understanding of the recipients and their reaction to the messages. It might be that behaviour can be changed in response to simple messages repeated frequently in many forums, but in developing countries there will a trade off between efficacy and cost: repeated home visits by friendly health workers may not be feasible on a large scale. It might also be the case that the desired changes in behaviour are not realistic for the individual or community because of economic, social, and cultural barriers. Interventions aimed at women must take into account their heavy workload in the home and field and their degree of influence within the household on decisions about child care, family planning, and health seeking behaviour.