HAWTHORNE Effect: A Methodological Problem in Growth Studies during Infancy A. K. BHALLA and Vijay KUMAR Advanced Centre for Delivery of Maternal and Child Health Care Department of Community Medicine, Post Graduate Institute of Medical Education and Research, Chandigarh Abstract Body weight and crown-heel length measurements were undertaken on 154 urban Punjabi infants by longitudinal and 721 comparable infants by cross-sectional methods. The growth attainments were comparable in the first 4/5 months, but after this age infants assessed by longitudinal method demonstrated significantly better growth attainments (p<0.05) as compared to those assessed by cross-sectional method. This pattern is explained on the basis of HAWTHORNE effect. It is recommended that the role of HAW- THORNE effect be considered while interpreting growth data from longitudinal studies. Keywords HAWTHORNE effect, Growth assessment Introduction The principal drawback of longitudinal studies are the time they take to complete and the small number of subjects that can usually be followed up. One of the ways to maintain adequate sample size in such studies is to generate a high level of involvement amongst children and their parents in the study. Unfortunately, the greater the involvement the more likely it is to affect the growth of children participating in the study. The phenomenon of greater involvement of parents in a longitudinal series yields to HAWTHORNS effect (BLALOCK, 1970), which results from subjects knowing that they are a part of study. According to JOHNSTON (1980), continuous and persis- tent health care advice to the parents (whose children are part of a longitudinal study) may become more aware of relationship between growth and need for proper environment. As a result they may change their children's diet etc. which may lead to improved general growth status of the children included in this study. No authentic information is available regarding this aspect of longitudinal studies and therefore the need to understand this phenomenon becomes obvious. One of the ways to understand this phenomenon could be a comparison between longitudinal and cross-sectional methods of studying human growth. In the present study an attempt has been made to quantify the magnitude of HAWTHORNE effect on growth pattern of Article No. 8518 Received June 24, 1985.
34 A. K. BHALLA and V. KUMAR Punjabi infants by comparison of weight and crown-heel length based on simultaneously collected longitudinal and cross-sectional data. Material and Methods The material for the present investigation is comprised of longitudinal and cross-sectional data for weight and crown-heel length, assembled on Punjabi* infants representing comparable socio-economic strata residing in Chandigarh. These infants were normal, healthy, full term and weighed over 2500g at birth. Infants with history of birth trauma, asphyxia or those suffering from any disease of central nervous system or from a significant congenital anomaly in the body were excluded from the study. Body measurements on two types of sample infants were taken by the same investigator (AKB) using standard techniques (WEINER & LOURIE, 1969) and instruments between 2 p. m. to 5 p. m, in the respective clinics. All the children included in the two series were weighed nude with help of a beam balance (Make; Detecto, capacity 16kg) upto the accuracy of 20g. Crown-heel length was measured with the help of an infantometer upto the accuracy of 0.1cm. The accuracy of these instruments was ensured throughout the study. Biological age of every child was determined on the basis of known date of birth, which in the case of cross-sectional sample subjects was verified from the certificate of birth registration or discharge slips of the hospital. Age and sexwise distribution of two types of samples is shown in Tables 1 and 2. * The parents of each child, who used Punjabi as their language for daily conversation and traced their ancestral place of origin to any place located in the present day Punjab province were considered Punjabis. Longitudinal Series Longitudinal examination of 154 infants (86 males and 68 females) was carried out in Growth Clinic and Neo-natal Follow-up Clinic attached respectively to Department of Community Medicine and Paediatrics of the Postgraduate Institute of Medical Education and Research, Chandigarh. Body measurements on each neonate were taken in the Labour room/maternity ward of the Nehru Hospital, Chandigarh within 24 hours of the birth. Whereas, subsequent monthly examination of each child between 1 to 12 months was conducted in the respective clinics with a tolerance of *3 days on the day of measurement. The infants who were not brought by the parents in respective clinics on the given dates, were followed-up at their homes after the working hours. Health status of every infant was periodically assessed by a medical doctor attached to the respective clinics. Advice on health care of each infant was also given to every mother, particularly regarding breast feeding, food supplementation and hygiene of feeding bottle etc. Cross-Sectional Series Cross-sectional data comprised of weight and crown-heel length measurements taken on a group of 721(365 males and 356 females) Punjabi infants. These infants were measured in the Well baby clinic attached to the Department of Paediatrics between 1 to 12 months of age at different monthly age levels with a tolerance of *3 days on the day of measurement. These infants were brought to the clinics either for the purpose of routine check-ups or immunisation. Measurements at birth had been taken on these infants also in the Maternity ward within 24
HAWTHORNE Effect 35 Table 1. Age-wise comparison of mean (*SD) values for weight (kg) and crown-heel length (cm) among Punjabi male infants of the longitudinal and cross-sectional series. * p<0.05 ** p<0.01 *** p<0.001 Table 2. Age-wise comparison of mean (*SD) values for weight (kg) and crown-heel length (cm) among Punjabi female infants of the longitudinal and cross-sectional series. * p<0.05 ** p<0.01 *** p<0.001
36 A. K. BHALLA and V. KUMARhours of the birth. The advice regarding breast feeding, food supplemen first year of life is shown in Tables 1-2 and Figs. 1-2. In their general course, curves plotted for longitudinal data indicate a greater smoothness as compared to those based on cross-sectional data. The differences in the mean longitudinal and crosssectional data values for weight and crownheel length are negligible at birth thus revealing an initial similarity in the basic nature of two types of data. This situation continues to be more or less similar upto 4 months barring male infants in whom such a situation for crown-heel length extends upto 5 months of age. Longitudinal curves for these two measurements run superior to those plotted on the basis of cross-sectional data throughout the remaining period of infancy. The difference in the mean values of weight and crown-heel length based on two types of data, at monthly age levels are found to be statistically significant (Tables 1 and 2) upto the age of 12 months with the infants representing longitudinal series more advanced in their growth pattern than those included in cross-sectional sample. Discussion This study shows that growth pattern of infants as measured by body weight and crown-heel length is similar during the first 4/5 months of life, irrespective of the method of evaluation whether cross-sectional or longitudinal. Thereafter, the growth pattern of children assessed by longitudinal method is significantly better (P<0.05) than that by the cross-sectional technique (Tables 1 and 2). No satisfactory explanation can be offered. Possible reasons to explain these differences include the additive manifestation of the HAWTHORNE effect. During the first few visits the impact is not measurable or not manifested because of the limited number
HAWTHORNE Effect 37of contacts with the fa a period of one year. Whereas such a situation was altogether absent in cross-sectional series where repeat examinations were not conducted. From a methodological point of view, HAWTHORNS effect appears to be an inbuilt and inevitable feature of longitudinal design of studying human growth. On one end it leads to the improvement of general growth status of children included in the longitudinal study solely because of the basic nature of the design. Whereas, on the other hand, it looks like harsh indictment of the longitudinal method, as infants with improved growth status stop representing the population from which they were originally drawn as sample. Above comments are not intended to convey that longitudinal design is not basic to the study of growth. Rather it is to emphasize that a longitudinal study should be initiated only after a careful consideration of all aspects of design, of which HAWTHORNS effect could be an important one.
38 A. K. BHALLA and V. KUMAR References BLALOCK, H. M., 1970: An Introduction to Socia velopment. In: Human Physical Growth and Maturation (eds. F. E. JOHNSTON, A. F. ROCHE, and Ch. SUSSANNE). Plenum Press, New York. WEINER, J. S. and J. A. LOURIE, 1969: Human Biology-A Guide to Field Methods. Blackwells Ltd., Oxford. Anil Kumar BHALLA Advanced Centre for Delivery of Maternal and Child Health Care Department of Community Medicine Post Graduate Institute of Medical Education & Research Sector-12, Chandigarh-160012, India