Sheet Music

A pilot study of a computer-assisted cell-phone interview (CACI) methodology to survey respondents in households without telephones about alcohol use

A pilot study of a computer-assisted cell-phone interview (CACI) methodology to survey respondents in households without telephones about alcohol use
of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A pilot study of a computer-assisted cell-phone interview (CACI)methodology to survey respondents in households without telephonesabout alcohol use CHRIS WILKINS, SALLY CASSWELL, HELEN MOEWAKA BARNES & MEGAN PLEDGER   Alcohol and Public Health Research Unit, University of Auckland, New Zealand  Abstract  An intrinsic drawback with the use of a computer-assisted telephone interview (CATI) survey methodology is that people who livein households without a connected landline telephone are excluded from the survey sample. This paper presents a pilot of the feasibility of a computer-assisted cell-phone interview (CACI) methodology designed to survey people living in households without a telephone about alcohol use and be compatible with a larger telephone based alcohol sample. The CACI method was found to bean efficient and cost competitive method to reach non-telephone households. Telephone ownership was found to make a differenceto the typical occasion amount of alcohol consumed, with respondents from households without telephones drinking significantly more than those with telephones even when consumption levels were controlled for socio-economic status. Although high levels of telephone ownership in the general population mean these differences may not have any impact on population alcohol measuresthey may be important in sub-populations where telephone ownership is lower.  [Wilkins C, Casswell S, Moewaka Barnes, H,Pledger M. A pilot study of a computer-assisted cell-phone interview (CACI) methodology to survey respondents inhouseholds without telephones about alcohol use.  Drug Alcohol Rev  2003;22:221–225] Key words:  computer-assisted cell-phone interview (CACI), coverage, methodology, surveys. Introduction Telephone interviewing has become the method of choice for many social and health surveys due to itsefficiency of contacting a random sample of thepopulation [1]. The use of a computer assistedtelephone interviewing (CATI) methodology providesimprovements in efficiency and quality control of datacollection. However, an intrinsic drawback with the useof telephone interviewing is that people who live inhouseholds without a connected landline telephone areexcluded from the survey sample. The impact this non-coverage bias has on the validity of survey findings isdetermined by two factors: (1) the magnitude of thedifference in the variable being measured between thosewho have a telephone and those that do not, and (2) theproportion of the households in the sample populationwho do not have a telephone [1,2]. In a number of studies of drug use and health behaviour, high levels of telephone ownership in the general population werefound to mitigate the potential effects of non-coveragebias due to non-telephone ownership [2–5]. In NewZealand, where telephone ownership is high by inter-national standards (96%), non-telephone ownershipwas found to have no statistically significant effect onpopulation measures of alcohol use collected using atelephone methodology [5].However,amongparticularsubgroupsofapopulationlevels of telephone ownership may be lower and/or theprevalenceofthehealthvariableunderstudyisknownorsuspected to vary greatly from the wider population. InNew Zealand, Maori (89%) and Pacific Island peoples(87%),low incomehouseholds (86%) and single-parenthouseholds (92%) all have lower levels of telephoneownership than the national average [6].The most common approach to gaining someunderstanding of the bias due to the exclusion of non-telephone households has been to supplement Received 8 July 2002; accepted for publication 22 January 2003.Chris Wilkins PhD, Alcohol and Public Health Research Unit, University of Auckland, PB 92019, Auckland, New Zealand; Sally CasswellProfessor, Alcohol and Public Health Research Unit, University of Auckland, PB 92019, Auckland, New Zealand; Helen Moewaka Barnes MPH,Alcohol and Public Health Research Unit, University of Auckland, PB 92019, Auckland, New Zealand; Megan Pledger PhD, Alcohol and PublicHealth Research Unit, University of Auckland, PB 92019, Auckland, New Zealand. Correspondence to Chris Wilkins, Centre for Social andHealth Outcomes Research and Evaluation (SHORE), Massey University  2 , Weelesley Street, Auckland, New Zealand, Tel: 64 93666136.Fax: 64 93665149; E-mail: Drug and Alcohol Review  (June 2003), 22, 221–225 ISSN 0959-5236 print/ISSN 1465-3362 online/03/020221–05 # Australian Professional Society on Alcohol and Other DrugsDOI: 10.1080/09595230100100651  telephone surveying with face-to-face surveying inhouseholds without telephones [1]. However, face-to-face surveying can be financially costly and time-consuming, undermining two of the key advantages of a telephone methodology. More importantly, face-to-face interviewing changes the mode of administration of the questionnaire (i.e. personal visit versus telephone),which brings into question the extent that the datacollected in the face-to-face interviews is compatiblewith the wider telephone interviewing, particularly withregard to sensitive topics such as alcohol and drug use.There remains a need for a supplementary methodol-ogy that is capable of reaching respondents in house-holds without telephones which is compatible with aCATI methodology and incorporates some of the costand time efficiencies of a CATI survey.This paper describes the pilot of a computer assistedcell-phone interview (CACI) methodology developedto sample respondents living in households withouttelephones as part of a larger CATI survey. The CACIpilot was conducted as a supplementary survey to aCATI survey of alcohol use in New Zealand’s largestcity: Auckland. The aim of this research was to pilot thefeasibility of the CACI procedure and to collect corealcohol measures to examine whether ownership of alandline telephone had an effect on measures of drinking. Methodology Mobile phones carried by fieldworkers were used toconduct alcohol interviews with people living in house-holds without telephones. The mobile phone permittedresidents in households without telephones to completethe same alcohol interview that was administered to alarger sample using a CATI methodology. The ageeligibility criteria, random selection procedure, inter-view protocols and questions asked in the CACIinterview were identical to the CATI interview.CACI fieldworkers canvassed neighbourhoods tolocate households without a connected land-linetelephone. On discovery of a household without atelephone the study was introduced to the residentsand an information sheet on the study was provided.The fieldworker then used their mobile phone to callthe CATI call centre and arranged for an interviewerto speak to the respondent. Once a CATI call centreinterviewer was on the line, the mobile phone waspassed on to the respondent who was then takenthrough the standard random selection process andinterview if convenient. The CATI interviewer couldmake an appointment for the fieldworker to return ata later date if it was not convenient to complete theinterview at that moment. Consistent with the largertelephone sample where telephone numbers werecalled back at least 10 times, fieldworkers returnedto an eligible household at least 10 times in order tocomplete an interview. Also, consistent with themonitoring standards of the CATI sample, 10% of the CACI mobile phone interviews were monitoredby a call centre supervisor at a separate telephonebooth to ensure high quality interviewing. CACIfieldworker’s recorded their progress around a neigh-bourhood and these logs were checked by the CACIsupervisor.The neighbourhoods to be canvassed were selectedusing the ‘Super Map’ database. This is a set of CD-ROMs published by Statistics New Zealand thatcontains demographic information from the 1996Census of Population. The information is availabledown to the level of a neighbourhood or mesh block. Amesh block is a geographical unit containing about 25households. For the purposes of the pilot the number of mesh blocks to be canvassed was restricted to a randomsample of 200 mesh blocks with levels of non-telephoneownership above 15% of households.The lower limit imposed on the level of non-telephone ownership in the mesh blocks sampled wasset to increase the likelihood that there was at least onehousehold in a mesh block without a telephone andensure the pilot remained within the budget allocatedfor the project. In a full CACI component no suchlimits would be imposed. The implication of the 15%+non-telephone ownership limit is the mesh blockscanvassed were from lower socio-economic householdsand these demographic characteristics are likely toconfound the alcohol measures. Because data were notcollected from all types of non-telephone households(i.e. ones living in mesh blocks with less than 15% non-telephone ownership) the data collected in the CACIpilot could not be used to measure the bias present withregards to the whole telephone sample. A full CACIcomponent would collect the data to make thesecalculations. The response rate achieved by the CACIpilot was 65%, and this compared favourably with thelarger CATI sample (61%).  Analysis The data were analysed to see whether having accessto a land-line telephone made any difference to thealcohol consumption measures reported. The mea-sures examined were abstention, typical occasionquantity and annual frequency of drinking. Compar-isons were made after adjusting for the demographiccharacteristics of the sample (Table 1). Randomeffects models were used with a random effect foreach area unit (a contiguous group of mesh blocks).Area units were known for the CACI sample and wereestimated by comparison with published lists for theCATI sample. The response typical occasion quantity was logged and frequency of consumption was 222  Chris Wilkins  et al .  transformed by power (0.3) to achieve normality. Theresponse of abstention was modelled with logisticregression. Phone ownership, the demographic vari-ables and their interactions with phone ownership,were entered into the models. Demographic variableswere discarded from the model if they had a  p -value of more than 0.10 with all interactions being consideredfor exclusion before main effects. Least squares meansfor the variable phone ownership, after being trans-formed back to the srcinal scale, are reported for eachmodel. Results are reported as significant if their  p -value is less than 5%. Results  Feasibility of the CACI  The CACI procedures worked well and interviews wereable to be conducted in a smooth and efficient manner.Several issues for improvement emerged from the pilotproject. Even on a pre-arranged cellular call plan, cellphone calls and interviews proved to be expensivecompared to landline interviews. It was found that itwas cheaper for CATI interviewers on a landline to callthe CACI fieldworkers’ cell phone rather than viceversa. As a result, CACI fieldworkers were directed toring the CATI centre and then once connected with aCATI interviewer get the interviewer to call them back to complete the interview.CACI fieldworkers required much less trainingthan would have been required to bring someoneup to the standard to conduct a face-to-face alcoholinterview. The recruitment and training of CACIfieldworkers could therefore focus solely on achievingsuccessful initial engagement with a prospectiverespondent and providing information on the study.CACI fieldworkers for the alcohol pilot wererecruited from the local neighbourhoods, and amongparticular ethnic groups, where the study was beenconducted, and this was likely to have assisted in thesuccess rate of the initial engagement and theresponse rate achieved.It was found that the CACI fieldworkers required aseparate supervisor who they could contact about any concerns. CATI supervisors were often too busy withthe regular CATI shift to answer routine queries. Aknowledgeable person available on a cell phone wassufficient for this task rather than a full-time dedicatedstaff member.Finally, there were the problems also associatedwith canvassing households for face-to-face inter-views, such as the presence of dogs and poor or noEnglish language skills. Fieldworkers were advisedto avoid houses where dogs roamed free in proper-ties. Differences in alcohol consumption related to phoneownership The raw percentages and transformed least squaresmean for the adjusted and unadjusted model arepresented in Table 2, along with the adjustmentvariables. For abstention, the unadjusted modelshows a significant difference in least squares meansbetween people with and without landline tele-phones. The adjusted model shows no significant Table 1.  Demographic features of the samples PhoneAvailable Not available1205 206Sex % %Male 46 38Female 54 62Age14–19 15 2020–29 20 3430–39 22 2440–65 43 23MaritalPartner 59 41Parents 18 19Other family members 8 23Others 8 8No other adult 7 8Ethnicity Pakeha 64 8Maori 10 35Pacific peoples 9 49Other 15 9Don’t know/refused 2 0EducationNo school qualifications 18 48School qualifications 28 31Certificates (teaching, trades etc) 21 9Degree or part-degree 25 4Other 6 5Don’t know/refused 1 3EmploymentFull time 58 33Part time 15 14Parenting 8 18Student 9 14Unemployed 3 16Others 6 6Total income 5 $10000 20 22$10000–  5 $20 000 7 23$20000–  5 $30 000 9 19$30000 and over 58 32Don’t know/refused 6 3*The category ‘Don’t know/refused’ has been omitted when itrounds to less than 1. CACI methodology in alcohol use telephone survey  223  difference. Ethnicity is very strongly confounded withphone ownership and caused this change in sig-nificance.For typical occasion quantity, for both the adjustedand unadjusted model, the least squares means aresignificantly different. Phone access made a differencein the typical quantity consumed for all employmentgroups, those without phone access drinking less,except for the unemployed who showed no difference.For annual frequency of consumption, for boththe adjusted and unadjusted model, the leastsquares means are not significantly different.However phone access causes differences in howoften different age groups and different educa-tional groups consume alcohol. Those in the 14– 19 and 20–29 age groups drink more often andthose with post-school certificates drinking lessoften if they do not have access to a land-linephone. Discussion The principal advantage of a CACI sample over aface-to-face supplementary sample is its compatibility with the CATI data collected. CACI interviews areconducted by the same trained CATI interviewerswho conduct the wider telephone interviews, and dueto the presence of supervision staff at the CATIlaboratory, to the same high standard as the CATIinterviews in households with telephones. The CACImethod provides a cost competitive means to offer acomputerized questionnaire to respondents living inhouseholds without telephones. However, CACIinterviews cost five times more than the standardCATI interview. (As a comparison McAuliffe  et al  .[1] indicates that face-to-face interviews cost betweenfour and 10 times more than a standard telephoneinterview)Telephone ownership was found to make adifference to the typical occasion amount of alcoholconsumed, with respondents from households withouttelephones drinking significantly more than those wholived in households with telephones even when levelsof consumption were controlled for the effects of socio-economic status. This measure of quantity of alcohol consumed during a drinking occasion is of importance to public health alcohol research given theimpact on intoxicated-related harm [7]. For absti-nence and frequency of alcohol consumed there wasno significant difference in measures between tele-phone and non-telephone households among lowersocio-economic households. Conclusion The pilot study demonstrated the CACI survey methodis a feasible approach to collect data on non-telephonehouseholds comparable with that collected using aCATI methodology. The demographic characteristicsof the respondents in the CACI pilot did not explain allthe differences found in alcohol use between telephoneand non telephone households. It is therefore notpossible to simply adjust a CATI sample using thesample’s socio-economic characteristics to account for Table 2.  Statistics on alcohol consumption PhoneAvailable Not available Adjustment variablesAbstainers (%) 1 Raw 21 36Unadjusted 3 21 31 phone, sex, ethnicity, education,Adjusted 3 26 24 employment and marital statusTypical occasion quantity (ml of absolute alcohol) 2 Raw (median) 38 128 phone, sex, age, ethnicity,Unadjusted 3 40 130 education, employment andAdjusted 3 53 110 employment by phoneFrequency of consumption (occasions per year) 2 Raw (median) 110 79 phone, sex, age, ethnicity,Unadjusted 3 111 90 education, total income, age by Adjusted 3 82 91 phone and education by phone 1 The figures are based on the entire sample. 2 The figures are based on the drinkers surveyed. 3 Least square means which were transformed back to the srcinal scale. 224  Chris Wilkins  et al .  non-coverage of non-telephone households. However,as noted at the beginning of this paper, in situationswhere the level of telephone ownership is high, as is thecase in the general population in New Zealand, theinclusion of non-telephone households has had noimpact on population measures collected. Yet, insubpopulations where telephone ownership is lower(i.e. low-income households, single-parent households,Maori and Pacific Island peoples in New Zealand) thedifferences related to non-telephone ownership bemore important and the inclusion of a CACI supple-mentary sample may be warranted. References [1] McAuliffe W, Geller S, La Brie R. Are telephone surveyssuitable for studying substance abuse? Cost, administration,coverage and response rate issues. J Drug Issues1998;28:455–82.[2] Anderson J, Nelson D, Wilson R. Telephone coverage andmeasurement of health risk indicators: data from the NationalHealth Interview Survey. Am J Public Health 1998;88:1392– 5.[3] Aquilino W. Telephone versus face-to-face interviewing forhousehold drug use surveys. Int J Addict 1992;27:71–91.[4] Ford E. Characteristics of survey participants with andwithout a telephone: findings from the third National Healthand Examination Survey. J Clin Epidemiol 1998;51:55–60.[5] Wyllie A, Black S, Zhang J, Casswell S. Sample frame bias intelephone-based research in New Zealand. NZ Statistician1994;29:40–53.[6] Statistics New Zealand. 2001 Census Snapshot 2 (Who hasAccess to the Internet?)—Media Release, 2002.[7] Klingemann H, Gmel, G, eds. Mapping the social con-sequences of alcohol consumption. London: Kluwer Aca-demic Publishers, 2001. CACI methodology in alcohol use telephone survey  225
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks