The impact of recall bias on the accuracy of dietary information

Van Zyl, Zoe (2014-04)

Thesis (MNutr)--Stellenbosch University, 2014.

Thesis

ENGLISH ABSTRACT: Background: A number of observational studies where information was obtained retrospectively have been used in the past to inform guidelines regarding allergy prevention. Studies looking at the causative/protective properties of infant dietary factors on diseases that occur later in life also rely on maternal recall many years later. It is unclear however what the effect of the recall bias was on the accuracy/quality of the information obtained. Objectives: The aim of the study was to determine the impact of recall bias 10 years retrospectively on the accuracy of dietary information in relation to breast feeding, weaning age and introduction of allergenic foods. A literature review was performed into studies assessing the accuracy of data obtained retrospectively and into studies using retrospective data to draw conclusions on the protective/causative factors of infant feeding in relation to food allergy. Methodology: An infant feeding questionnaire was developed from some of the same questions that were asked by mothers recruited into the FAIR study, a prospective birth cohort on the Isle of Wight. Families had been recruited and followed up since 2001/2002 and data has been gathered when the mothers were 36 weeks pregnant, and then when their child was 3, 6, 9 months and 1 and 2 years old. Mothers were asked in 2012, when their children were 10 years of age, to complete this questionnaire. Agreement of answers was computed using Kappa coefficients, Spearman’s correlation and percentage agreement. Results: One hundred and twenty five mothers completed the questionnaire. There was substantial agreement for recall of whether mothers breast fed, the duration of EBF and breast feeding 10 years earlier (k = 0.79, r = 0.70 and r = 0.84 respectively). Seven per cent (n = 9) of mothers however who did breast feed reported not to have. Eighty four per cent (n = 103) of mothers recorded correctly whether their child had a bottle of formula milk in hospital. Ninety four per cent (n = 116) of mothers recalled accurately that their child had received formula milk at some stage of their infancy. The exact age at which formula milk was first given to their child was answered accurately (r = 0.63). The brand of formula milk provided was poorly recalled. Answers to when mothers first introduced solid foods into their child’s diet were not accurate (r = 0.16). The age of introduction of peanuts was the only food allergen that mothers recalled accurately for when they first introduced this into their child’s diet (86% correct answers). Recall of whether peanuts were consumed during pregnancy was accurate after two years (k = 0.64) but not after 8 years (k = 0.39). Conclusion: The study highlights the importance of possible recall bias of infant feeding practices by mothers over a period of 10 years. Recall related to breast feeding and formula feeding were accurately recorded for, but not for age of introduction of solid foods and introduction of allergenic foods. Studies relying on maternal recall of weaning questions need to be cautious.

AFRIKAANSE OPSOMMING: Agtergrond: ’n Aantal waarnemingstudies waarin inligting op retrospektiewe wyse of terugwerkend bekom is, is in die verlede gebruik om riglyne oor die voorkoming van allergie neer te lê. Studies oor die veroorsakende/beskermende kenmerke wat kindervoedingsfaktore op latere siektes het, steun verder op die herinneringe wat die moeder baie jare later kan oproep. Dit is egter onduidelik watter uitwerking hierdie oproepvooroordeel op die akkuraatheid/gehalte van die versamelde inligting het. Oogmerke: Die oogmerk met die studie was om die impak te bepaal wat oproepvooroordeel met terugwerkende effek van 10 jaar op die akkuraatheid van voedingsinligting oor borsvoeding, speenouderdom en die insluiting van allergeniese voedselsoorte uitoefen. ’n Literatuuroorsig was onderneem van studies wat die akkuraatheid evalueer van data wat retrospektief bekom is, asook studies wat retrospektiewe data gebruik om gevolgtrekkings oor die beskermende/veroorsakende kenmerke van kindervoeding met betrekking tot voedselallergie te maak. Metodologie: ’n Kindervoedingsvraelys is saamgestel vanaf sommige van die vrae wat aan gewerfde moeders voorheen in die FAIR-studie, ’n voornemende geboortekohort op die eiland Wight, gestel is. Gesinne is in 2001/2002 gewerf en opgevolg, en data is versamel toe die moeders 36 weke swanger was; en weer toe hulle kinders die ouderdom van 3, 6, 9 maande en 1 en 2 jaar bereik het. In 2012, toe hulle kinders 10 jaar oud was, is die moeders weer versoek om hierdie vraelys in te vul. Ooreenstemming tussen antwoorde is bepaal deur Kappa koeffisiënte, Spearman korrelasies en persentasie ooreenstemming. Resultate: Eenhonderd vyf-en-twintig moeders het die vraelys ingevul. Daar was beduidende ooreenkoms in die moeders se oproep oor die vraag of hulle borsvoeding gegee het, hoe lank eksklusiewe borsvoeding (EBV) geduur het, asook borsvoeding 10 jaar vantevore (k = 0.79, r = 0.70 en r = 0.84 onderskeidelik). Sewe persent (n = 9) van die moeders wat wel borsvoeding gegee het, het egter geantwoord dat hulle dit nie gegee het nie. Vier-en-tagtig persent (n = 103) van die moeders het akkuraat geantwoord op die vraag of hulle kinders bottelvoeding met ’n melkformule in die hospitaal ontvang het. Vier-en-negentig persent (n =116) van die moeders kon akkuraat oproep dat hulle kinders in ’n sekere stadium van hulle kindertyd melkformule ontvang het. Die vraag oor presies hoe oud die kinders was toe hulle die eerste maal melkformule ontvang het, is akkuraat beantwoord (r = 0.63). Die handelsnaam van die melkformule kon nie goed herroep word nie. Antwoorde oor wanneer moeders die eerste maal vaste voedsel by hulle kinders se dieet ingesluit het, was nie baie akkuraat nie (r = 0.16). Die ouderdom waarop grondboontjies ingesluit is, was die enigste antwoord wat moeders akkuraat kon oproep (86% korrekte antwoorde) op die vraag wanneer hulle die eerste maal ’n voedselallergeen by hulle kinders se dieet ingesluit het. Die antwoord op die vraag of hulle tydens hul swangerskap grondboontjies geëet het, was akkuraat na twee jaar (k = 0.64), maar nie na agt jaar (k = 0.39) nie. Gevolgtrekking: Die studie onderstreep die belang van moontlike oproepvooroordeel rakende kindervoedingspraktyke by moeders oor ’n tydperk van 10 jaar. Die oproep oor borsvoeding en formulevoeding is korrek aangedui, maar nie vir die ouderdom waarop vaste voedselsoorte en allergeniese voedselsoorte ingesluit is nie. Studies wat op moederoproep oor speningsvrae staatmaak, moet omsigtig gedoen word.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/86429
This item appears in the following collections: