Comparison of fieldworker interview and a pictorial diary method for recording morbidity of infants in semi-urban slums
© Thomas et al.; licensee BioMed Central. 2015
Received: 19 May 2014
Accepted: 7 January 2015
Published: 31 January 2015
Cohort studies conducted in low-income countries generally use trained fieldworkers for collecting data on home visits. In industrialised countries, researchers use less resource intensive methods, such as self-administered structured questionnaires or symptom diaries. This study compared and assessed the reliability of the data on diarrhoea, fever and cough/cold in children as obtained by a pictorial diary maintained by the mother and collected separately by a fieldworker.
A sample of 205 children was randomly selected from an ongoing birth cohort study. Pictorial diaries were distributed weekly to mothers of study children who were asked to maintain a record of morbidity for four weeks. We compared the reliability and completeness of the data on diarrhoea, fever and cough/cold obtained by the two methods.
Of 205 participants, 186 (91%) ever made a record in the diary and 62 (30%) mothers maintained the diary for all 28 days. The prevalence-adjusted bias-adjusted kappa statistics for diarrhoea, fever, cough/cold and for a healthy child were 92%, 79%, 35% and 35% respectively.
Diary recording was incomplete in the majority of households. When recorded, the morbidity data by the pictorial diary method for acute illnesses were reliable. Strategies are needed to address behavioural factors affecting maternal recording such that field studies can obtain accurate morbidity measurements with limited resources.
KeywordsReliability Children Morbidity measurements Slum India Fieldworker Pictorial diary Kappa Agreement
Quality of data is critically important in any study. Several aspects of epidemiological data collection such as completeness, clarity, interviewer’s skill and education level of the responder determine the quality of data . Cohort studies measure exposure factors at different time points  to evaluate the association between exposure and disease. Cohort studies in low-income countries largely use trained fieldworkers to make frequent visits to a participant’s home for routine surveillance, as opposed to industrialised countries where less resource intensive approaches, such as mailing self-administered structured questionnaires or maintaining a diary to record day to day morbidity are possible. Diarrhoea surveillance programs often employ both interviews and diary methods to record morbidity and obtain data on severity of illness .
The fieldworker interview has many advantages like reliability, validity and high response rate but incurs costs in terms of training, travelling and time which can have considerable impact on study size and the need for resources . Diary methods and self-administered questionnaires are more effective and beneficial in populations with high literacy levels and are not always recommended in low literacy level settings [5,6], but where used, are suggested for daily events [5,7-9].
We hypothesised that morbidity data collected through a pictorial diary method maintained by the mother would be as good as fieldworker records on a structured questionnaire used at home visits. The study compared the morbidity data recorded by a pictorial diary method against the data collected by the fieldworkers and assessed reliability of data collection for diarrhoea, fever and cough/cold.
Study setting, participants and data collection
Definitions used by the fieldworkers and mothers in assessing illness among study children
Time interval for a new episode
Three or more watery stools per day or a change in number or consistency reported by the mother and which she considers indicative of diarrhoea
48 hrs after cessation of the previous episode
Increased temperature of the body as perceived by primary caregiver
72 hrs after cessation of the previous episode
Cough/runny nose with or without fever
72 hrs after cessation of the previous episode
We estimated that approximately 200 subjects would be needed to provide 90% power to test the null hypothesis of kappa = 0.4 versus the alternate hypothesis kappa = 0.6 at 0.05 significance level (2 sided). A total of 205 subjects were selected by simple random sampling method using the cohort study database as a sampling frame and the pictorial diaries were distributed to their mothers by a separate set of fieldworkers not involved in the main cohort study. The purpose of the study and the definitions used (e.g. diarrhoea consists of 3 or more looser than normal stools in a 24 hour period) were explained to the mothers, with instructions to mark the child as healthy if there was no cough/cold, fever or diarrhoea. For 4 weeks, the fieldworkers visited the study mothers once a week, collected the previous week’s pictorial diary and handed over the next week’s pictorial diary.
The main cohort study fieldworkers were masked about the identity of pictorial diary participants and they continued their biweekly surveillance per the main study protocol. The morbidity data collected for the 205 children during the study period by the fieldworkers were extracted from the original study database for analysis. The fieldworker collected data were subjected to a 10% random recheck by a field supervisor through the duration of the study.
Double data entry was done and verified using Epi-Info 3.5.1 (CDC, GA, USA) for data collected in the diaries. The socio-demographic characteristics and surveillance data were extracted from the main study database. SPSS 16 (SPSS Inc., IL, USA) and STATA 10 (StataCorp, TX, USA) software were used for analysis.
A test of association between baseline characteristics and completeness of the diary was performed using Pearson’s chi-squared test. McNemar’s chi-squared test was performed to compare correlated proportion of reported days of illness by both methods; p < 0.05 was considered to be statistically significant. The reliability of number of child illness days was calculated using kappa statistics and prevalence-adjusted bias-adjusted kappa (PABAK) . In order to interpret kappa meaningfully it is important to report prevalence and bias along with kappa and provide adjusted kappa. A higher prevalence index results in lower kappa whereas higher bias index results in higher kappa and PABAK is used to adjust these paradoxes and the interpretation of the strength of agreement is the same as kappa [13,14]. For agreement on episodes of illness, percent agreement was used .
Child-days of follow up by the fieldworker and by the mothers with the pictorial diary
Child-days (Number of subjects)
Number of child-days of follow up
Number of child-days completed for all 28 days
Comparison of missing child-days of follow up
Number of child-days completely missed
Number of child-days recorded
Number of child-days partially recorded
Characteristics of the mothers and children who participated in the study
N = 205 n (%)
Age of the mother
More than 24 years
Less than or equal to 24 years
Number of years of mother’s education
More than 5 years
Less than or equal to 5 years
Socio economic status
Middle and High
Sex of the child
Birth weight of the child (N = 203*)
Greater than or equal to 2500 grams
Less than 2500 grams
Birth order of the child
Third or later born
First or second
Age of the child
More than 6 months
Less than or equal to 6 months
We did not find any significant difference among the mothers who completed (n = 62) and who did not complete (n = 124) the diary in age, education of the mother, socioeconomic status (SES) of the family, birth order of the child and family size.
Illness reporting, inter-rater agreement and episodes
Illness reported by fieldworker interview and by a pictorial diary maintained by the mother for 4 weeks in 186 children for whom data were available for any length of time by both methods
Number of days reported
Cough and cold
>1 symptom reported
Reliability of two methods using Kappa statistics for observations reported by both methods (3897 observations in 186 children)
Percent agreement examining paired observations of episodes of illnesses between fieldworker and diary data
Biweekly morbidity surveillance data collected by fieldworkers were used to test the reliability of data reported daily in a pictorial calendar by the mothers. While 91% of the mothers had made at least one entry, only 30% of the mothers completed the diary for the entire 28 day period. No baseline difference could be identified for mothers who completed or did not complete the diaries.
A community based clinical trial in Australia which aimed at a high level of completion of a health diary over 68 weeks and a study in Canada to test the validity and feasibility of diary data collected by parents at certain times after vaccination and up to 21 days reported very high levels of completeness [16,17]. Diary methods are generally recommended for populations with adequate literacy levels [6,18], but a pictorial diary does not require literacy. Pictorial diary methods have been used in health utilisation, health expenditure and morbidity surveys [19-22]. Although maintaining a diary for long periods could result in fatigue and attrition , some studies suggest that encouraging the participants to improve compliance would result in improved reporting both in terms of accuracy and completion [7,9].
Among the reported illnesses, diarrhoea and fever showed considerable overlap while cough/cold had more reports by fieldworkers than in the diaries. One explanation could be under-reporting by mothers of common minor illnesses, as up to 7 respiratory illnesses per child per year have been previously reported in this setting .
Mothers reporting more diarrhoeal and fever episodes and fewer cough/colds (Table 6) could be attributed to sporadic marking in the diary for specific episodes, especially for longer duration episodes recorded by the field workers. This resulted in more than one episode being recorded instead of a single long episode if there were two or more missing days for diarrhoea and three or more missing days for fever. Since coughs/colds needed to be recorded for five or more days to count as an episode, the lack of marking in the diary resulted in fewer episodes than compared to fieldworkers’ records.
Reliability assessment (Table 5) shows that for diarrhoea and fever the proportion observed, chance agreement and prevalence index are high which resulted in lower kappa and when adjusted for the two paradoxes  bias and prevalence, the PABAK shows substantial agreement  for diarrhoea and fever respectively. An Argentine study reported poor kappa agreement of mother’s perception captured through a written questionnaire about their overweight and obese children compared to body mass index z-score  whereas a Bangladesh study showed that 60% of the mothers correctly identified malnutrition in their children . A cross sectional study in Kenya reported that a pictorial method identified malnutrition better than verbal description though both methods under-estimated malnutrition when compared to formal anthropometric measurements . The published literature on the use of pictorial methods in low-income settings therefore indicates that the pictorial diary method could be effective in recording illnesses if high rates of compliance could be achieved. Although only 30% of mothers recorded 4 weeks of data, their data were reliable and comparable to conventionally collected data, which suggests that even though diaries are vulnerable to attrition and fatigue they do not compromise the quality of data, as has been reported previously [7,8,23].
In order to avoid response bias which is one of the limitations of diary methods , the field workers did not insist that the mothers should complete the diaries and just collected the weekly pictorial diaries and issued blank ones. This study therefore assessed spontaneous diary completion, but reasons for not completing the diaries were not collected, which is the main limitation of this study.
There was a high rate of attrition in pictorial diary use by mothers over a 4 week period. Where collected, the morbidity data recorded by pictorial diary methods for acute illnesses was reliable when compared to fieldworker collected data. Reasons why mothers did not complete diaries were not collected. Future studies should examine behavioural factors affecting motivation to complete diaries as a possible strategy to improve data collection in resource limited studies.
We would like to thank Mr. Ethiraj, Ms. Anjugadevi, Ms. Muthulakshmi, Ms. Magimai, Mr. Sivaprasath, Ms. Geetha and Mr. Ilaiyaraja for their help with the data collection, Mr. Kaviarasu and Mr. Shanmugam for their help with data entry. We are also grateful to the mothers and their families for their participation and support. This study was supported by National Institutes of Health grant NIAID RO1 AI072222. DK was supported by FIC training grant D43 TW007392 (GK).
- Whitney CW, Lind BK, Wahl PW. Quality assurance and quality control in longitudinal studies. Epidemiol Rev. 1998;20(1):71–80.PubMedGoogle Scholar
- White E, Hunt JR, Casso D. Exposure measurement in cohort studies: the challenges of prospective data collection. Epidemiol Rev. 1998;20(1):43–56.PubMedGoogle Scholar
- Lewis K. Vesikari clinical severity scoring system manual. Seatle: PATH; 2011.Google Scholar
- Burcu A. A comparison of two data collection methods: Interviews and questionnaires. Hacettepe Univ J Educ. 2000;18:1–10.Google Scholar
- Bruijnzeels MA, van der Wouden JC, Foets M, Prins A, van den Heuvel WJA. Validity and accuracy of interview and diary data on children’s medical utilisation in The Netherlands. J Epidemiol Community Health. 1998;52(1):65–9.PubMedPubMed CentralGoogle Scholar
- Bruijnzeels MA, Foets M, van der Wouden JC, Prins A, van den Heuvel WJA. Measuring morbidity of children in the community: a comparison of interview and diary data. Int J Epidemiol. 1998;27(1):96–100.PubMedGoogle Scholar
- Wiseman V, Conteh L, Matovu F. Using diaries to collect data in resource-poor settings: questions on design and implementation. Health Policy Plan. 2005;20(6):394–404.PubMedGoogle Scholar
- Verbrugge LM. Health diaries. Med Care. 1980;18(1):73–95.PubMedGoogle Scholar
- Sullivan LM, Dukes KA, Harris L, Dittus RS, Greenfield S, Kaplan SH. A comparison of various methods of collecting self-reported health outcomes data among low-income and minority patients. Med Care. 1995;33(4):183–94.Google Scholar
- Sarkar R, Sivarathinaswamy P, Thangaraj B, Sindhu KN, Ajjampur SS, Muliyil J, et al. Burden of childhood diseases and malnutrition in a semi-urban slum in southern India. BMC Public Health. 2013;13:87.PubMedPubMed CentralGoogle Scholar
- Kattula D, Sarkar R, Sivarathinaswamy P, Velusamy V, Venugopal S, Naumova EN, et al. The first 1000 days of life: prenatal and postnatal risk factors for morbidity and growth in a birth cohort in southern India. BMJ Open. 2014;4:e005404. doi:10.1136/bmjopen-2014-005404.PubMedPubMed CentralGoogle Scholar
- Byrt T, Bishop J, Carlin JB. Bias, Prevalence and Kappa. J Clin Epidemiol. 1993;46(5):423–9.PubMedGoogle Scholar
- Sim J, Wright CR. The Kappa Statistic in Reliability Studies: Use, Interpretation, and Sample Size Requirements. Phys Ther. 2005;85(3):257–68.PubMedGoogle Scholar
- Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159–74.PubMedGoogle Scholar
- Gordis L. Epidemiology. 4th ed. Philadelphia: Saunders Elsevier; 2005.Google Scholar
- Hellard ME, Sinclair MI, Forbes AB, Fairley CK. Methods used to maintain a high level participant involvement in a clinical trial. J Epidemiol Community Health. 2001;55:348–51.PubMedPubMed CentralGoogle Scholar
- Freeman TR, Stewart M, Birtwhistle R, Fisher DC. Health diaries for monitoring events following immunization. Can J Public Health. 2000;91(6):426–30.PubMedGoogle Scholar
- Das J, Hammer J, Sánchez-Paramo C. The impact of recall periods on reported morbidity and health seeking behavior. Washington DC: The World Bank; 2011.Google Scholar
- Ansah EK, Powell-Jackson T. Can we trust measures of healthcare utilization from household surveys? BMC Public Health. 2013;13(853):1–5.Google Scholar
- Boeke CE, Mora-Plazas M, Forero Y, Villamor E. Intestinal protozoan infections in relation to nutritional status and gastrointestinal morbidity in Colombian school children. J Trop Pediatr. 2010;56(5):299–306.PubMedGoogle Scholar
- Wright JA, Gundry SW, Conroy RM, Wood D, Du Preez M, Ferro-Luzzi A, et al. Defining episodes of diarrhoea: results from a three-country study in Sub-Saharan Africa. J Health Popul Nutr. 2006;24(1):8–16.PubMedGoogle Scholar
- Mwangome MK, Fegan G, Prentice AM, Berkely JA. Maternal perception of malnutrition among infants using verbal and pictorial methods in Kenya. Public Health Nutr 2014:1–8. doi:10.1017/s1368980014001074Google Scholar
- Verbrugge LM. Sensitization and fatigue in health diaries. In: Proceedings of the American Statistical Association (Survey Research Methods Section). Washington DC: American Statistical Association; 1980. p. 666–71.Google Scholar
- Feinstein AR, Cicchetti DV. High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol. 1990;43(6):543–9.PubMedGoogle Scholar
- Hirschler V, Gonzalez C, Talgham S, Jadzinsky M. Do mothers of overweight Argentinean preschool children perceive them as such? Pediatr Diabetes. 2006;7:201–4.PubMedGoogle Scholar
- Roy SK, Rahman MM, Mitra AK, Ali M, Alam AN, Akbar MS. Can mothers identify malnutrition in their children? Health Policy Plan. 1993;8(2):143–9.Google Scholar
- Howe LD, Galobardes B, Matijasevich A, Gordon D, Johnston D, Onwujekwe O, et al. Measuring socio-economic position for epidemiological studies in low-and middle- income countries: a methods of measurement in epidemiology paper. Int J Epidemiol. 2012;41(3):871–86.PubMedPubMed CentralGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.