WPC1M 2B R ZV#|P HP LaserJet 4HPLAS4.PRSx  @\ubX@USUK 3'3'Standard6&6&StandardHPLAS4.PRSx   2 XX"m^3ETggEEEgt3E39gggggggggg99ttt[~rEP~rr~•~E9E`gE[g[g[Egg99g9ggggEP9gggg[c)coE3EE"EEECEEEEEEdEg9YYYYYY~Y~Y~Y~YC8C8C8C8ddddddddddYggdddsd[[[[[[[g~[~[~[~[ggggggggE8E9E9E9oPg~9~9~9~8~9gvggggEEErPkNrPrP~9~9~9gggggg•g~[Y~[g~9gErP~9gggggNH3]gE[ggggg9@gFdddgF%C[[gEEggzzdddwrrE[dF"^E[p[[[E[ELLLz[k[L[[zzz[LLL[kLz7[E[[.[[[Y[[[[[[[LwwwwwwwwwwYJYJYJYJwzzzzzzzzzzz[J[L[L[LkLLLJL[[[khkkLLLzwzL[kLN`E|[zLU^^1Yzz[[    [z^"m^3ErggΫEEEgu3E39ggggggggggEEuuug~Pg•~rΕE9ExgEgr[r[Egr9Er9rgrr[PErggg[Q-QkE3EE"EEECEEEEEEdEr9dddddΕYYYYYN8N8N8N8oddddooooddrgdddzoggg[[[[r[[[[ggggggrrP8P9P9P9rgr999N9rrrggΕ[[[rPoNrPrPEEErrrrrrΕg[Y[r9r[rPEggrgrNH3ogEgggggg>DgFdddgF9CgggEEggzzddd~rrEgdF"^E[[[[E[EL[[k[L[[zz[L[Lzk[zl<l[E[[.[[[Y[[[[[[[LwwwwwhJhJhJhJzzzzzzzzkJkLkLkLLLLhLzzzkhkk[[[zwzLzk[N`E[R[^^LY[[    [^29E  |"m^gCdddgC9CgggEEggzzdddsrrEgdC2e& k  ( * %"m^)2>>g`)))>F)">>>>>>>>>>""FFF7rYRRYKEYY)0YKnYYEYREKYYuYYK)"):>)7>7>7)>>"">"`>>>>)0">>Y>>7;;C))){)))())))))<)>"W5W5W5W5W5nRR5L5L5L5L5(!(!(!(!YY>WK7K7K7K7Y>Y>Y>Y>Y>Y>Y>Y>)!)")")"XC0Y>K"K"K"L!K"Y>YGY>Y>Y>Y>nYR)R)R)E0@/E0E0K"K"K"Y>Y>Y>Y>Y>Y>uYY>K7N5K7Y>K"Y>R)E0K"Y>Y>Y>Y>Y>N+WW8>)7>>>>i>"&kk>*<<<>k*(77>{))xx>>II[x><q*"xxxxWWxxxWWkkxxx>$jjjjjq5jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjzzzzvvvvvvvvvv^^jjjjjjssjj[[[[}}jjggggjj``}`[[[[7+J[jjCourierCourier BoldTimes New RomanTimes New Roman BoldTimes New Roman ItalicTimes New Roman Bold ItalicCG TimesCG Times BoldCG Times ItalicCG Times Bold Italic2L  &`1 6 A"4|J~ ^;C]ddCCCdCCCCddddddddddCCdzN`zoCCCddCdoYoYFdo8Co8odooYNCodddYdddddd"dddCddddddddo8dddddYYYYYN8N8N8N8oddddooooddoddddzodddYYYYoYYYYddddddooN8N8N8N8r`o888N8oooddYYYoNoNoNoNCCCoooooodYYYo8oYoNCddodoCddCCCWxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxNdddCdUUddddddFddddFCCssd44ddzzddd~ooCsdF"dsd9dCCzCddoddCdYds`zUwdddCCCCzozoYNYYYN8YooYdYzzdzddYYzozzzzNY88YYYzYzzzzCCdddddddzzzzzzzzzzzzzzzzzzzNNNNNNNYYYYYYYYYYYYYYYYYYYY888888888888YYYYYYYYYYYYYYYYYYYzzzzzzzzzzzzzzzzzzzzCzNzNddYCx \   pxtllXphh<X\H|?xxx,2x6X@`7X@D?xxx,)x `7X IYE,B& \  P6 P LIYE,%T 4  p  7iC3,ƒXi\  P6XP  7nC3,%4Xn4  pX N@{N<,M{\  P6P M@N<,%4  pZ 5hC3,:Xh*f9 xr XX  6jC3,Q3Xj9 xOX W!?(,,h?\  P6hP8wC;,mXw PE37XP  V"G($,mhG PE37hP7zC;,vc!Xz_ pi7X>6uC;,cXu&_ x7XX|4xC;,Xx*0 x]7X"4|J~ ^;C`ddCCCdCCCCddddddddddCCdzzzsCYozzdozzooCCCddCddYdY8dd88Y8ddddNN8dYYYNYdYd3dd"dddCddddddddd8zdzdzdzdzdYzYzYzYzYC8C8C8C8dddddddddoYzddddoYdzdzdzdzdYYYYdzYzYzYzYddddddddC8C8C8C8dYYo8o8o8o8o8dzddddzNzNzNdNdNdNdNo8o8o8ddddddoYoNoNoNdo8dzNdNo8oYoYdddCddCCCWxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxNdddCd]]ddddddFddddFCCddd88ddzzdddkddCddF"ddd9dCCzCdzdoddCdYds]zUwdYYCCCCzzzozoYzNoYdYC8YooYdYzzdzddoYoYzzozzzzNd88YYYzYzzzzCCdddddddzzzzzzzzzzzzzzzzzzzNNNNNNNdddddddddddddddddddd888888888888YYYYYYYYYYYYYYYYYYYzzzzzzzzzzzzzzzzzzzzCszzCozCoddYCx \   pxtllXphh<X\H|"4|J~ ^;NhddCCCdCCCCddddddddddCCdzNdzzozzzCCCddCddYdYFdo88d8odddNN8oYdYNdddddd"dddCddddddddd8dddddYYYYYN8N8N8N8oddddoooozYddddzYdzddddYYYYdYYYYddddddooN8N8N8N8kddz8z8z8zFz8oooddNNNoNoNoNoNz8z8z8oooooozYzNzNzNdz8oNoNz8zYzYddoCddCCCWxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxNdddNdYYddddddCddddCCCkkd88ddzzdddsssCkdC"d~d9dCCzCddoddCdYds`zUwdddCCCCzozoYNzYYYN8YooYdYzzdzddzYzYzozzzzNY88YYYzYzzzzCCdddddddzzzzzzzzzzzzzzzzzzzNNNNNNNYYYYYYYYYYYYYYYYYYYY888888888888YYYYYYYYYYYYYYYYYYYzzzzzzzzzzzzzzzzzzzzCzNzzNzddYCx \   pxtllXphh<X\H|2/M # \  P6B& P##Xi\  P6ƒXP#@VIII - @    a] #{\  P6MP#CHAPTER VIII #Xi\  P6ƒXP#  a  #Xi\  P6ƒXP##{\  P6MP#MONITORING TOBACCO USE: PREVALENCE AND SURVEY  ag INFORMATION  X #Xi\  P6ƒXP#INTRODUCTION Prevalence of tobacco use in the population is one important measure of the magnitude of the tobacco problem. Additionally, with information about the prevalence of tobacco use in different subgroups of the population, it is possible to identify who are the high risk groups for tobacco use. Such information is helpful for planning effective health education and tobacco intervention programmes for the appropriate target groups. Knowledge about prevalence levels also strengthens the position of tobacco control advocates in lobbying for tobacco control measures. If an intervention or educational programme is to be launched in the community, then the prevalence data can be used to establish the baseline for evaluating the effectiveness of such a programme. Repeated periodic prevalence surveys in the same population group would provide particularly useful information on trends concerning tobacco use behaviour. Given that tobacco is used in a variety of forms, when measuring prevalence, it is essential to specify the type of tobacco being used, especially when more than one type of tobacco use is common. For example, in India, one should report prevalence of overall tobacco consumption as well as bidi smoking, betelquid chewing, etc. *p-++Ԍ X ԙ SOME BASIC DEFINITIONS XFor purposes of standardization it is important that terms and concepts be defined in a concise way. This will permit global monitoring of the tobacco epidemic and comparisons between countries. To facilitate this, a series of key definitions are given below."   XAny population can be divided into two categories, smokers and nonsmokers."  X A:` ` A smoker  is someone who, at the time of the survey, smokes any tobacco product either daily or occasionally."`  X ` `  Smokers  may be further classified into two categories:"`  X` XA1:X` ` A daily smoker  is someone who smokes any tobacco product at least once a day, (with the exception that people who smoke every day, but not on days of religious fasting, are still classified as daily smokers)."`  X XA2:X` ` An occasional smoker  is someone who smokes, but not every day."` ` ` Occasional smokers include:  X ` ` A2 i): %Reducers people who used to smoke daily but now do not smoke everyday."  XQ XX` ` A2 ii):-Continuing occasional people who have never smoked daily, but who have smoked 100 or more cigarettes (or the equivalent amount of tobacco), and now smoke occasionally."  X ` ` A2 iii):` -Experimenters people who have smoked less than 100 cigarettes (or the equivalent amount of tobacco), and now smoke occasionally."  X?& XB:X` `  A   nonsmoker  is someone, who, at the time of the survey, does not smoke at all. "`  X( XX` `  Nonsmokers can be classified into three categories:"` Z*p-++Ԍ X XB1:X` `  Ex smokers  are people who were formerly daily smokers, but currently do not smoke at all."`  X XB2:X` `  Never smokers  are those who either (i) have never smoked at all, or (ii) have never been daily smokers and have smoked less than 100 cigarettes (or the equivalent amount of tobacco) in their lifetime."`  X B3:` `  Exoccasional smokers  are those who were formerly occasional, but never daily smokers and who smoked 100 or more cigarettes (or the equivalent amount of tobacco) in their lifetime."` XX` ` These definitions can be used to classify the population according to their lifetime smoking status. In particular:"`  X XC:X` `  Ever smokers  are defined as those who have ever smoked at least 100 cigarettes (or the equivalent amount of tobacco) in their lifetime."` XX` ` A specific subcategory of interest are those who have smoked, or now smoke, everyday."`  X5 C1:  ` `  Ever daily smokers  are defined as persons who are currently daily smokers, reducers or are exsmokers."` A convenient schema for illustrating the relationship between these various categories is given in Figure VIII.1. p-++  X  Figure VIII.1 :  %Classification of the population according to current and previous  X smoking status " Y ddx !ddxbo ' Y     >Current smoking status z  Previous smoking status 3Daily 3(A1) IOccasionally M(A2) ^Nonsmoker b(B)z Z 5  X7  Daily smokers Daily smokers (A1) Reducers (A2 i)   Ex smokers (B1)Z ,  Never smoked daily but smoked 100 or more cigarettes5 G35 Continuing occasional smokers (A2 ii)5 Exoccasional smokers (B3),    Never smoked daily and never smoked as much as 100 cigarettes J j4J Experimenters (A2 iii)J  Never smokers (B2) s 5 Never smoked at all  j4 jN s J Some common categories of smoking status for individuals can then be readily constructed from this schema as follows: Smokers = daily smokers (A1) + Occasional smokers (A2 (iiii)) Ever smokers = daily smokers (A1) + occasional smokers (A2 (iiii)) ` `  %+ ex smokers (B1) + exoccasional smokers (B3) Ever daily smokers = daily smokers (A1) + reducers (A2 i) + exsmokers (B1) XEx smokers = exdaily smokers (B1)" In a subsequent section of this chapter, details are provided about the relationship between these categories of smoking status and the responses to the seven core questions on individual smoking behaviour which are proposed in this chapter to assess smoking practices from population surveys.)p-++Ԍ X X Smokeless tobacco use : The foregoing definitions apply equally well to smokeless tobacco use (e.g. snuff, betel quid chewing, other chewing tobacco). "   X X Tobacco user  is a person who is either a smoker or a smokeless tobacco user, or both."  X2  THE CONCEPT OF PREVALENCE    X  Prevalence of smoking is defined as the "proportion (usually expressed as percentage) of the population who are smokers (both daily and occasional) at a point in time". In other words: O !ddxbo ' Addxb O seJ  L L '.Number of smokers in the population being surveyed Prevalence of =   x 100 smokers (in %)L L '.6TT=Total size of the survey population  L L '. (smokers and nonsmokers) eb  X  Example : A survey of 400 men and 500 women aged 15 and over yielded 200 smokers among men and 50 women who smoked. Male smoking prevalence is therefore: ` ` 200   x 100 = 50% ` ` 400 Female smoking prevalence is similarly calculated as: ` ` 50   x 100 = 10% ` ` 500  X@% Note of caution: Prevalence estimates should always be calculated for males and females separately. If a total prevalence estimate is also required (i.e. for both sexes combined), then it may be calculated as the sum of the male and female smokers divided by the sum of the male and female sample populations provided that there has not been oversampling of either sex in the sample population. In other words, the sample*p-++ composition should accurately reflect the population sex composition for this procedure to be valid. In this example, the combined prevalence (M + F) is calculated as: ` ` 200 + 50  20 x 100 = 27.8% ` ` 400 + 500  It is not appropriate to calculate total prevalence as the simple arithmetic average of male and female prevalences without weighting the sexspecific prevalences to reflect the actual sex composition of the population.  Thus, for calculating prevalence of smoking, two numbers are required: a numerator and a denominator. The numerator is the number of smokers and the denominator is the number of persons from which smokers are derived (and includes smokers as well as nonsmokers). The population for which the prevalence is being determined is reflected in the denominator. For example, if the denominator represents a specific community, the figure would refer to the prevalence of smoking in that community. If the numerator and denominator refer to a specific subgroup, then the prevalence refers accordingly to that particular subgroup. Prevalence should be calculated separately forr different subgroups of the population according to age, sex and other relevant sociodemographic characteristics such as educational level, ethnicity and linguistic groups. In each case, the number of smokers in the subgroup should be divided by the total number of persons (smokers plus nonsmokers) in that subgroup of the population to calculate subgroupspecific prevalence. If diverse types of tobacco use are present, it is possible to calculate prevalence of different methods of tobacco consumption from the same data set. When doing so,  X% the denominator would remain the same, only the numerator would change. For  X' example, the numerator should be the number of bidi smokers when estimating the prevalence of bidi smoking. )p-++Ԍ  X]  The foregoing definitions of prevalence should be used to calculate the  X prevalence of different categories of tobacco use. For example, ` `  %-hh4<CNumber of daily smokers in the surveyed ` `  %-hh4<Cpopulation Prevalence of daily smoking =hh4   x 100 ` `  %-hh4<CTotal size of the surveyed population ` `  %-(daily smokers, plus occasional smokers plus nonsmokers)  ` `  %-hh4<CNumber of smokeless tobacco users ` `  %-hh4<C(daily and occasional) Prevalence of smokeless tobacco use =    x 100  \ M M  % ,UU4; Total size of the surveyed population (daily M M  % ,UU4;Csmokeless tobacco users, occasional and nonusers)  It is useful to monitor the process of quitting smoking. For exsmokers, two prevalence measures can be calculated. The first of these defines the percentage of exsmokers in the entire population and is calculated as follows: M M  % ,UU4Number of exsmokers in the surveyed population Prevalence of exsmokers =   x 100 M M  % ,UU4Total size of the surveyed population (ever smokers M M  % ,UU4and never smokers) One problem with the above measure is that it does not give a clear indication of quitting behaviour among those who may be classified as ever daily smokers. To provide information on the  \ percentage of ever daily smokers who are exsmokers, a new measure, the prevalence of cessation, should be used. This is defined as follows: M M  % , Number of exsmokers in the surveyed population Prevalence of cessation =   x 100 M M  % , Number of ever daily smokers in the surveyed population   X6& Other terms, such as the quit rate, the quit index, or the quit ratio, have also been used to describe this or a similar measure. P*p-++Ԍ X ԙ QUESTIONS TO ASSESS TOBACCO USE For determining prevalence, the smoking status of each individual in the sample needs to be established. To do this, a series of questions, or a questionnaire which can be administered to the sample, needs to be developed. The questionnaire should be prepared carefully because the accuracy, completeness and usefulness of the data is largely dependent on the questionnaire.  X The following core questions are recommended for assessing smoking status. In a questionnaire, these questions should be accompanied by other questions seeking sociodemographic information on the respondents (as a minimum age, sex and socioeconomic status). p-++ @ Addxb addx]  @ a(a(]   Xb  Q 1 XHave you ever smoked? (Y/N). If no, stop interview/questionnaire here.Ƒ!  X  Q 2 XHave you ever smoked at least 100 cigarettes or the equivalent amount of tobacco in your lifetime? (Y/N).Ƒ!  X3  Q 3 XHave you ever smoked daily? (Y/N)Ƒ!  X  Q 4 XDo you now smoke daily, occasionally, or not at all?Ƒ!  XJ (D (Go to Q 5) , O (Go to Q 7) , Not at all (Go to Q 7) )  X  Q 5 XOn average, what number of the following items do/did you smoke per day?Ƒ! ........ manufactured cigarettes ........ handrolled cigarettes ........ bidis ........ pipefuls of tobacco ........ cigars/cheroots/cigarillos ........ goza/hookah  X  Q 6 XHow many years have you smoked/did you smoke daily?Ƒ!  X  Q 7 XHow long has it been since you last smoked daily?Ƒ!   less than one month   one month or longer but less than six months   six months or longer but less than one year   one year or longer but less than five years   five years or longer but less than ten years   10 years or longer. a(]) p-++Ԍ  W]  Smokeless tobacco use  Apart from smoking, tobacco is used in smokeless forms. The most common is oral use which may consist of chewing a tobacco product or applying the product on the  X gums. It may also include nasal use. For assessing smokeless tobacco use, the following additional set of questions are recommended.  p-++ @ addx]  ddx]  @ '']  X  Q 1 XHave you ever used smokeless tobacco? (Y/N). If no, stop interview/questionnaire here.Ƒ!   X  Q 2 XHave you ever use smokeless tobacco at least 100 times in your lifetime (Y/N). Ƒ!  X  Q 3 XHave you ever used smokeless tobacco daily? (Y/N)Ƒ!  X  Q 4 XDo you now use smokeless tobacco daily, occasionally, or not at all?Ƒ!  X (D Go to Q 5) , O (Go to Q 7), Not at all (Go to Q 7) )  X  Q 5 XOn average, what number of the following items do/did you use per day?Ƒ! ........ snuff (oral use) X........ snuff (nasal use)Ƒ! ........ chewing tobacco ........ betel quid X(there is a great diversity of smokeless tobacco products used in different countries and this list should be modified accordingly).Ƒ!  X  Q 6 XHow many years have you used/did you use smokeless tobacco daily?Ƒ!  XI  Q 7 XHow long has it been since you last used smokeless tobacco daily?Ƒ!   less than one month   one month or longer but less than six months   six months or longer but less than one year   one year or longer but less than five years   five years or longer but less than ten years   10 years or longer. '] ) p-++Ԍ X ԙ Explanatory Notes for Core Questions  X  The following notes are intended to provide further clarification on these core questions. These notes should assist interviewers and those responsible for data processing, in addition to those responsible for overall survey management.  X.  Q 1 XThose who have smoked very few (even one) cigarette(s) in their lifetime should still answer "yes" to this question.%"  XE  Q 2 X100 cigarettes is approximately the same as smoking one cigarette per day for three to four months, or occasional smoking for about one year.%"  X\  Q 3 XIf desired, the qualification "for at least six months" could be added to preserve comparability with previous WHO definitions.%"  Xs  Q 4 XThis refers to current smoking status at the time of the survey.%"  X-  Q 5 XThis list of items should be adapted to suit local tobacco use patterns. Some of these items may not be used and therefore should be excluded, but others not listed might be common and therefore should be included.%" XIf some items are smoked, but the average is less than one per day, use the number zero. For former daily smokers, the average should refer to the time when they were daily smokers.%"  X  Q 6,7 XThe actual daily amount smoked should be recorded. Precoded categories such as 110 per day, 1120 per day etc., should be avoided. Should only be asked of ever daily smokers (i.e. those who answer "yes" to Q 3 and/or who answer "daily" to Q 4). Enter number of years. If less than one year, code as zero. Do not count time periods when the person was not smoking.%" C* p-++Ԍ X  Q 7 XThis question should only be asked of exdaily smokers (i.e. those who answer "yes" to Q 3 and "occasionally" or "not at all" to Q 4). Enter number of years. If less than one year, code as zero. Do not count time periods when the person was not smoking. Respondents who have stopped smoking as recently as one day prior to the survey should still be classified as exsmokers (with a duration of less than one month).%"  X  ASSESSING TOBACCO USE IN SPECIAL POPULATIONS  XE  Women of reproductive ages, pregnant women Because there are additional risks of smoking which are unique to women of reproductive age and pregnant women, there is often a need for more detailed knowledge of tobacco use patterns among these populations. The standard questions to elicit smoking prevalence information should still be asked with additional questions on intentions to quit, knowledge of health hazards and opinions about tobacco and health issues, as appropriate.  X  Health professionals: Practitioners, teachers and students Health professionals are highly respected in their communities. People frequently turn to them for advice and example on health matters. For this reason, exemplary behaviour, knowledge, practice and opinion of health professionals, both in practice and in training, are very important elements in advancing tobacco control policies. Surveys of health professionals, in practice and in training, should include standard questions on smoking prevalence, as well as additional questions on knowledge of the health effects of tobacco, together with questions about curricular content and health professional practice with respect to counselling patients about cessation of tobacco use. C* p-++Ԍ X  Assessing adolescent tobacco use  X  Surveys of adults will contain more items on quitting than on initiation, because adults are more likely to quit smoking than are adolescents. Conversely, surveys of adolescents should focus on the process of initiation, because most people start smoking as adolescents. Adolescents smokers appear to be more likely than adult smokers to misclassify themselves as nonsmokers. This bias appears to be more common in household surveys than in school surveys. Confidentiality is more likely to be compromised in household interviews. Steps to improve the confidentiality of the household setting, for example, by administering the survey instrument as a selfadministered questionnaire, are likely to improve the validity of survey estimates of smoking prevalence. Smoking status of adolescents should never be assessed by proxy reports, because parents, who are likely to provide answers, are often not aware that their children smoke. The questions and definitions used for adults can be applied to adolescents, with some modifications. Since adolescents are in the process of establishing a daily pattern of smoking behaviour, a much smaller percentage of adolescents than adults who have smoked in the previous month will be daily smokers. Among persons who ever smoked daily, the process of progressing from first trying a cigarette to smoking daily usually takes about two to three years. The following questions are recommended:  X[  To assess ever smoking , ask, "Have you ever tried or experimented with cigarette smoking, even one or two puffs?"  Xr#  Age of first use : If the person has tried smoking, ask "How old were you when you first tried a cigarette?"  X'  Number of lifetime cigarettes smoked : If the person has tried smoking, ask "How many cigarettes have you smoked in you entire life?" C*p-++Ԍ X  Frequency of smoking : For those who have ever tried smoking ask whether they now smoke daily, less than daily but at least once a week, less than weekly but at least once a month, less than monthly, or not at all.  Xt  Likelihood of smoking : A series of questions are recommended to determine if an adolescent has a firm resolve not to smoke. For example, adolescents who report that they will not try a cigarette soon, who would definitely not smoke a cigarette if one was offered by one of their close friends, and who state that they are definitely not going to smoke at any time during the next 12 months, can be classified as not susceptible to smoke. Less definitive negative responses, positive responses, refusals, don't knows, and missing responses would result in the classification of the person as being susceptible to smoke. Among adolescents who have never tried smoking, those who are susceptible to smoke are more likely than those who are not to subsequently try smoking and to progress to daily smoking.  Xs  Minors access to cigarettes : Adolescents should be asked how they obtain their cigarettes. The questions should be phrased wither in general terms (i.e. "How do you usually obtain the cigarettes you smoke?") or in more specific terms (i.e., "How did you obtain the last cigarette you smoked?"). Response items should include: purchased in a store, purchased from a vending machine, from a friend or sibling, from one of my parents, had a friend purchase them for me, and stole them from a store. This information will be extremely useful to policy makers in countries where the sale of tobacco products to minors is illegal.  X  SELECTING KEY QUESTIONS While it is strongly recommended to always ask these seven core questions to assess smoking behaviour, in some case, it might only be possible to include one or two questions (for example, in a population census or a large nationallyrepresentative survey). C*p-++Ԍ X If only one question is permitted, it is recommended to ask the following:  X A i:"Do you now smoke daily, occasionally or not at all?"  This question will permit the assessment of the main categories of current smoking status, but nothing else. If this is to be asked in a population where smokeless tobacco use is significant (e.g. prevalence of 5% or more), the question may be amended as follows:  X A ii:"Do you smoke, use smokeless tobacco only, use both or none, and how  V frequently (daily, occasionally or not at all)?"  X If there is an opportunity to ask two questions, then it is recommend to either ask an additional question about daily consumption of tobacco or to ask about previous history of tobacco use. These questions might be formulated as follows: B i:X(for daily users) "How many of the following items do you smoke, chew or apply per day?%" ......... manufactured cigarettes ......... handrolled cigarettes ......... bidis ......... pipefuls of tobacco ......... betel quids ......... snuff" B ii:X(for all respondents) "Have you ever smoked daily, occasionally or not at all?%" ......... daily ......... 100 or more cigarettes, but never daily ......... not at all, or less than 100 cigarettes in your lifetime"  X(  #Xi\  P6ƒXP#  By combining responses to questions A i and B ii, the prevalence of all major categories of smoking status (current and lifetime) can be determined.O*p-++Ԍ X ԙ DATA COLLECTION A questionnaire can contain two types of questions: closeended and openended. In closeended questions, the number of possible responses is limited the respondent must choose from a group of predetermined responses. Openended questions are those where the respondent is not given responses to choose from, and replies are not categorised in advance. For example, the question "why did you start smoking"? or "why did you stop smoking"? would generally fall in the openended category. Openended questions need to be coded before any statistical analysis can be done are thus are time consuming and more difficult to analyze, especially for large samples. Therefore, it is recommended that the questionnaire be limited to closeended questions wherever possible. Since it is almost impossible to reinterview the same individuals for additional information, the questionnaire should include all the information that may be required at the time of analysis. Sufficient details, such as age, sex, education level, age at smoking initiation or duration of use should be ascertained. Also, the interviewer should check the responses for errors and inconsistencies before leaving the interviewee. It is also recommended that only key questions that will contribute toward the objectives of the survey should be included. It may be argued that there is very little additional cost of collecting information on other topics of interest while conducting a prevalence survey. However, inclusion of unrelated items may dilute the overall quality of the data collection, and would also require a longer time commitment from each interviewee. Therefore, it is recommended that the temptation to include extra unrelated questions in the questionnaire be avoided. To ensure this, it is helpful to plan beforehand what analysis will be done. Each question in a questionnaire should be precisely defined and exact criteria specified whenever necessary. Even a question that sounds very simple and straightforward may mean different things to different persons. For example the question "Are you a smoker?" seems clear. But this does not indicate whether the respondent is a daily or occasional smoker. To avoid problems and confusion it is bestC*p-++ to clearly define the criteria (as recommended in this chapter) and include them in an instruction manual. Further, it is strongly recommended to use the standardized questions proposed above, taking into account the cultural context of the population to be surveyed. The instruction manual should detail all the steps that are to be taken by the interviewer before and during the interview. It should also contain an explanation of each question in the questionnaire, the definitions, the criteria, the range of possible answers, and examples of unacceptable answers. The instruction manual is important not only for the interviewer but it is also helpful as a reference for the investigator, especially while writing the report. In the interviewadministered situation, the interviewers should receive training on the questionnaire to ensure uniform use of the questionnaire. Pretesting, an important part of a questionnaire's development process, is performed by administering the questionnaire to selected individuals who have similar characteristics to the target population. The purpose of pretesting is to discover potential problems or identify questions that can be misinterpreted, so that the questionnaire may be revised before it is administered to the larger population. Pretesting does not require a very large sample but may need to be repeated several times until the final questionnaire is adopted. A convenience sample of about 50 people covering all important subgroups might be used for this purpose. There are several methods of obtaining responses to survey questions. With a selfadministered questionnaire, the questionnaires are given, or mailed out to prospective interviewees, who fill them out by themselves. This is a common method in developed countries; however, in many developing countries, due to a high rate of illiteracy, this method has only limited applicability. It may be feasible and effective only for specific segments of the population that are better educated, such as doctors and other health professionals, office workers, or college students. The strengths and weaknesses of different approaches to completing questionnaires are summarized in Table VIII.1. C*p-++ԌIn developing countries, the most common method of obtaining survey data is through a trained interviewer. Generally, the questions are asked during a facetoface interview, in which the interviewer asks the interviewee all the questions, one by one, and records the answers. Explanations, if necessary, are given as included in the questionnaire and instruction manual. Leading questions, or questions that provide a specific answer, or a category of answers, as being more desirable than others, should be avoided. (For example: "How many cigarettes do you smoke, about 10 per day?"). If possible answers are to be supplied as part of a question, then all possible answers should be placed before the interviewee. It is also feasible to ask the questions via telephone, particularly in developed countries, where telephones are commonplace. In fact, random digit dialling is a common method of sample selection in developed countries. However, in most developing countries, this method is not widely applicable because telephone ownership is limited. If the selected interviewee is not available, it may be tempting to ask the questions of a close relative or a friend of the interviewee concerning the tobacco use of the originally selected person. For a survey of prevalence of tobacco use, such a procedure is not recommended. When distributing selfadministered questionnaires by mail, special efforts should be made to obtain a high response rate. The questionnaire should be typographically neat and should give the impression that it will be easy to complete. A friendly covering letter should also be included. It is also important to send out reminders as necessary. The first reminder should go out shortly after the original mailing. Subsequent reminders might be accompanied by an extra copy of the questionnaire. Up to three or four reminders might be required to achieve the desired response rate (which should be at least 70%). C*p-++ԌSAMPLING CONSIDERATIONS The most important part of a prevalence survey is the selection of the sample that forms the "denominator" or the population about which smoking behaviour is being assessed. The prevalence data refer to prevalence in the population group which form the denominator. In other words, the prevalence refers to the population, or segment of the population, that is represented by the denominator. Sometimes the denominator in a prevalence survey does not represent an easily identifiable or representative segment of the population, and in such instances, the prevalence has only limited value. In fact, all else being equal, the usefulness of a prevalence estimates is directly proportional to the representativeness of the denominator in terms of the population that is being studied. Probability sampling Since is it not possible to ask everyone in a population about their smoking status, a sample is selected which should yield a prevalence level similar to that in the whole population. The most appropriate method for selecting such a sample is  XD probability sampling, in which each individual in the population has a known, specific probability of being included in the sample. Such a procedure requires prior information about the population and careful planning for selecting the sample. Ideally, one would have a list of individuals (or their identification numbers) in the population, and the required number of interviewees could be selected by a simple random sampling procedure. Random sampling does not mean haphazard or "at will" sampling. This is a specific, scientific sampling procedure that provides a known probability for each member of the population to be included in the sample, and is generally accomplished by the use of random numbers. Lists of random numbers are available in several publications, and most computers can also generate random numbers.F*p-++ԌThe selection procedure is simple. Suppose there are 9000 individuals numbered from 0001 to 9000. Since a computer generated random number lies between 0 and 1, multiply it by 10 000, discard the fractional part and select the individual corresponding to the integer part in the sample. For example, if the random number generated by the computer is 0.4632717, then multiplying by 10 000 yields 4632.717. Disregarding the fractional part (i.e. number after the decimal) gives the number 4632. The individual corresponding to this number in the survey population is then chosen for inclusion in the sample. If the integer is more than 9000 disregard the number and repeat the procedure until the desired sample size is obtained. Duplicate selections should be discarded as well. Generally, it is not possible to have a list of all individuals in the population, but it may be possible to get a list according to groups of houses or wards (small geographical units defined for administrative purposes). For example, the census office can often supply a list of wards in a city with a detailed breakdown of the population of each ward. In such cases, rather than selecting individuals by random sampling, a "ward" can be considered as the sampling unit and the desired number of wards selected. Every person in selected wards may be included in the sample (see Appendix VIII.1,  X  Case Study I ), or a random sample of individuals may be chosen from each randomly  X selected ward (two stage sampling) (see Appendix VIII.1, Case Study II ). An example of a multistage sampling schema for a national household survey is given in Appendix VIII.2. The disadvantages of this form of sampling are that it is expensive and time consuming to search for each selected sampling unit (or individual). In addition, this form of sampling requires prior information about the population, and such prior information may not always be available in developing countries. Although probability sampling may not be feasible for an individual investigator, especially on a large or national scale, this approach is often used by various government agencies for their own purposes. It may be possible to persuade an agency to collect information on tobacco use if they are collecting related information. For example, inC*p-++ India, the National Sample Survey collected information on household expenditure on tobacco as part of their household expenditure survey. In Malaysia, the Dental Department in the Ministry of Health conducted a survey of tobacco habits as part of a  X survey on oral precancerous lesions (see Appendix VIII.1, Case Study II ).  X  Nonprobability sampling  X  Probability sampling is the only scientifically valid way of assessing prevalence in the community. Every effort should be made to obtain prevalence data in this way. However, if resources are not available to do so, then other less desirable methods which  X have been used in the past might be considered . These are described below. It is extremely important that whenever these methods are used, the limitations of the results should be clearly described in the survey report.  X  Opportunistic sampling . Due to the difficulties of probability sampling, it might be tempting to interview easily accessible individuals, such as in a crowded market place. Obviously, such a sample would not be representative of the population of the entire city since everyone is not equally likely to be at that place. Additionally, the interviewer might only approach individuals who seem friendly, and among them many would refuse. Thus, without being able to identify the segment of the population of the city that is represented by the sample, the prevalence estimate would be of little use.  X This procedure is therefore not recommended.  X`  Hospital sampling. If resources are very limited, one method would be to interview persons coming into a hospital, with the permission and cooperation of hospital  X" authorities (see Appendix VIII.1, Case Study III ). It would not be appropriate to interview patients because many more smokers have health problems and are more likely to be patients compared with nonsmokers. However, the friends and relatives that accompany the patients may be interviewed. If it is not possible to interview each person who is eligible, a selection procedure should be decided in advance. For example, depending upon the arrival rate and capacity for interviewing, the accompanying person of every third or fourth patient may be interviewed.J*p-++ԌIf this approach is used, careful attention must be given to the choice of hospitals such as large and small units, public and private institutions in order to cover as wide a population as possible. Hospital sampling requires very little resources because only the services of one or two interviewers would be required. The chances of cooperation would be high in this setting, since both the patients and their accompanying relatives or friends would typically have time to answer a few questions on tobacco use.  XE This method is not equivalent to probability sampling since everyone in the population is not equally likely to come to the hospital.  X_  Occupational groups . Prevalence can also be relatively easily estimated for different occupations, professions or employment status in a particular office or industry.  X Prevalence obtained in this way is not representative of that of the population. Whenever these data are presented, it should be made clear that they refer only to the occupation group in question. One advantage is that since the group is already welldefined, enough information should be available for simple random sampling, if a sample is required. Once the cooperation of the management is obtained, there should be few obstacles to getting responses from individual members. If all members of the group are literate, costs can be reduced by using selfadministered questionnaires rather than facetoface interviews. The prevalence obtained using this procedure could be fairly representative of the selected group. Since the group is highly specialised, extrapolation to the general population is not advisable. Despite these limitations, it may be preferable to study such welldefined and stable groups, rather than samples from the general population, if long X$ term followup is the main goal. (See Appendix VIII.1, Case Study IV ). K*p-++Ԍ X  Sample size The most important factor in establishing the reliability or precision of the estimated prevalence is the size of the sample. The larger the sample size, the more reliable the estimated prevalence, provided the sample is a probability sample. If the sample size is small, the chance that the prevalence estimated from the sample is close to the true prevalence is reduced, and the prevalence estimate from the survey could be substantially different from the true prevalence. The reliability of the prevalence estimate in a simple random sample is proportional to the square root of the sample size. Therefore, reliability (precision) increases somewhat slowly with an increase in sample size. As a guide, the reliability of the sample estimate of prevalence should be within five percentage points of the true prevalence figure. Table VIII.2 gives details of the required sample size depending on the prevalence estimate and the precision required. An example of sample size calculation is also given in the table. A sample size of less than 1000 could yield unreliable estimates especially if subgroup estimates are required and participation is not very high. For most populationbased prevalence surveys, a sample size of 20004000 adults would typically be required to yield reasonably reliable estimates for population subgroups of interest. In addition to a single estimate for the whole population, and separate estimates for men and women, prevalence for different subgroups should also be sought. In such cases, the reliability of the estimate would be determined by the size of the subgroup sample. Therefore, when determining the overall sample size, the reliability of estimates within desired subgroups need also to be taken into consideration. As discussed earlier, these statistical considerations are valid and meaningful only if the sample is a probability sample. For other methods, where there is no precise idea of the population from which the sample was drawn, confidence intervals and other statistical procedures have little meaning and the prevalence estimates should be viewed with great caution.C*p-++Ԍ X ԙ Repeating Surveys for Monitoring Trends in Tobacco use To measure changes in tobacco use, over time, repeated surveys are necessary using the same data collection techniques to ensure comparability of data. It is essential to use the same set of core questions asked in the same way and processed and reported in a similar fashion for the same subgroups as previously used. For monitoring changes in the prevalence of tobacco use in the population, each new survey should select an independent sample to that used earlier. If survey methods need to be changed for subsequent surveys, every effort should be made to measure the impact of the change on estimated prevalence. For example, each of the survey methods should be applied to appropriate samples at the same time in order to evaluate the effect of different survey methods.  X As a minimum, repeat surveys should be conducted every five years at the same time of year. If resources permit, surveys should be conducted more frequently,  X  optimally each year . If resources are limited, it is strongly recommended that rather than conducting large prevalence surveys at longer intervals, it would be preferable to conduct more frequent prevalence surveys on a modest scale at shorter intervals. One reason for doing this is that data from successive surveys can be pooled to make estimates for subgroups. This is particularly recommended when the epidemic is changing rapidly.  X[  Data preparation, analysis and presentation For comparability of data, it is necessary to use both standardized questionnaires for data collection and standardized protocols for the processing of data and the presentation of results, e.g. in tables. C*p-++Ԍ X  "Cleaningup"/editing of data The first task in the analysis phase is to carry out the cleaningup or editing of the data. This implies checking the data for consistency, and checking the responses for completeness and accuracy. It is important to monitor the number of missing answers and to check for inconsistencies in responses (e.g. if a certain number of cigarettes per day have been reported for nonsmokers). In principle, some of these inconsistencies can be eliminated by carrying out the following procedures: i)Xby checking data against the questionnaire (to exclude eventual data entry errors when an inconsistency appears in a data file)%" ii)Xby entering the data twice into the computer%" iii)Xby establishing "rules" for recording one of the variables in question (for example: if a respondent has answered "no" on Q 1, "yes" on Q 2, and responded 10 on the number of cigarettes in Q 5, then inconsistencies may be handled by defining the answers to all three questions concerned as missing thus excluding the respondent from the reporting of the variables in question. However, this situation would not have arisen if all responses had been checked and corrected by the interviewer at the time of the interview.%"  W  Computation of response rates For computing a response rate for a survey, the following data need to be recorded:  X.&  N :Xtotal number of persons in the original random sample%"   X(  D :Xnumber of persons who could not be approached because they have died or emigrated. These people are then excluded from the denominator%"E*p-++Ԍ X ԙ M :Xnumber of persons eligible for the study and available for answering the  X] questionnaire ( M = N D). Note that the eligibility criteria should have been defined before the full work begins, and should be strictly adhered to. %"  Xt  R :Xnumber of persons who actually replied to the questionnaire%"  X M M  % ,UU4 ;C]]JR R  X Then the survey response rate will be C]]J  x 100  X M M  % ,UU4 ;C]]JRN D   WF  Prevalence of different categories of tobacco habits It is important to realize that the number of respondents is not the same for each question (e.g., nonsmokers will not be asked to answer the question on the number of cigarettes smoked daily). Therefore the number of respondents may or may not coincide with what is the relevant basis for calculation of prevalence percentages. The following two examples may help to clarify this point. i)X"daily smokers" are individuals identified as those who have answered "daily" on%"  X XQ 4. The prevalence of this category is NOT the proportion of those who answered daily out of the three possible answers to this question, because this distribution does not refer to the whole survey population, only to those who have answered "yes" to Q 1. Those who have answered "no" on Q 1 have to be added to the denominator in order to get the correct prevalence figure for daily smokers.%" ii)X"nonsmokers" are identified as the sum of those having answered "no" on Q 1 and those having answered "not at all" to Q 4. The appropriate denominator for calculating their prevalence would then be the total number of respondents to Q1.%" For the purposes of data processing, the following correspondence rules table provide a guide to the responses (to the core questions) which should be used to classify respondents into various categories of smoking. These guidelines assume that all inconsistencies in the data have been identified and resolved.D*p-++  X    X CATEGORY  % ,UU4 ;CCORRESPONDING RESPONSES TO M M  % ,UU4;CQUESTIONS ON THE CORE M M  % ,UU4;CQUESTIONNAIRE:eea h h h h h h h h h h h  X  Daily smokers % ,UU4;Q 4: Daily  X_ ԩ of cigarettes only % ,UU4; Q 5 : some number >0 on the line(s) for h h h h h h h h h h h h h h h ;manufactured or handrolled cigarettes and 0 on M M  % ,UU4;all other lines  X ԩ of both cigarettes and otherUU4; Q 5 : some number >0 both on the line(s) for forms of tobacco (mixed smokers);cigarettes and on at least one other line  X ԩ only of tobacco forms other than UU4;Q 5 : zero on the line(s) for cigarettes and some cigarettesM M  % ,UU4;number >0 on at least one other line  X  Occasional smokers ,UU4;Q 4: Occasionally who are former daily smokersUU4;Combination of Q 4: Occasionally and Q.3: Yes (reducers) who have never smoked dailyUU4;Combination of Q 4: Occasionally and Q 3: No but smoked 100 or more cigarettes;and Q 2: Yes (continuing occasionals) who have never smoked dailyUU4;Combination of Q 4: Occasionally and Q 3: No and smoked < 100 cigarettesUU4;and Q 2: No (experimenters)  X Nonsmokers  % ,UU4; Q 1: No or Q 4: Not at all who are former daily smokersUU4;Combination of Q 4: Not at all and Q 3: Yes (Exsmokers) who have never smoked daily but;Combination of Q 4: Not at all, and Q 3: No, have smoked 100 or moreUU4;and Q.2: Yes cigarettes (exoccasionals)  X  ԩ who have never smoked at allUU4; EITHER Q 1: No; OR combination of Q 2: No, or smoked less than 100 cigarettes;Q 3: No and Q 4: Not at all and never daily (never smokers) _________________________________________________________________________ Ever smokers % ,UU4;Q 1: Yes Ever daily smokers % ,UU4;Q 3: Yes _________________________________________________________________________ NB: For all the above categories prevalence percentages should be based on the total number of respondents to Q 1.  *p-++ԌIn preparing computer programmes for the processing of data, these combinations can readily be translated into standard, "IF ... THEN ..." statements whereby the desired "defined variables" can be created.   These same general principles will also be applicable to the analysis and presentation of data on the use of smokeless tobacco. When assessing patterns of total tobacco use it should be kept in mind that there may be a certain overlap between smokers and users of smokeless tobacco. The prevalence of a category such as "daily tobacco users" (including smokers and/or users of smokeless tobacco) can therefore not be derived simply by adding the prevalence of "daily smokers" and "daily users of smokeless tobacco", since the "combined users" would then be counted twice. All categories of tobacco users should be analyzed and presented separately for males and females. For each sex, data should be presented by age. Ten yearagegroups starting with 1524 years will generally be sufficient. Broader age ranges could also be used, especially when the sample size is relatively small. If this is necessary, then a recommended age breakdown is 1524 years, 2544 years, 4564 years, and 65+ years and over. The table below provides an example of the calculation and presentation of prevalence estimates from a hypothetical set of survey data. [p-++  X  Table VIII.3  X]  Hypothetical example of the presentation of survey data on smoking behaviour  X   Age Group (years) % ,UU41524; 2544 4564 65+  YAll ages SexM M  % ,UU4M F M F M F M F M F  Desired sample size % ,UU4244 250 189 213 467 479 263 362 1163 1304 Actual number of respondents UU4244 243 189 212 457 476 257 350 1147 1281 Missing answers (data) ,UU4 0 7 0 1 10 3R 6 12 16 23  Prevalence estimates (calculated on the basis of actual number of respondents, ignoring missing data) Daily smokers (%) %20 34UU4 26 37C 30 31R 19 15 24 29 Occasional smokers(%)19 18UU4 14 9C 10 6R 4 3 12 8 Nonsmokers (%) %61 48UU4 60 54C 60 63R 77 82 64 63  Total (%)M M  %100 100UU4100 100C100 100R100 100 100 100  When reporting prevalence estimates from a survey, care should be taken not to exaggerate the precision of the figures. At most, one decimal place should be reported.  Reporting of results from surveys of adolescent smoking requires a different approach, because the rates of tobacco use change rapidly during adolescence. The  X frequency of smoking categories discussed in the Assessing adolescent tobacco use section should be used. Ideally, these should be reported for single years of age for each year of age from 10 to 19, with summary reporting for the groups 1014 and 1519. Unfortunately, the costs of obtaining statistically reliable data for single years of age may be prohibitive. If this is the case, then data could be reported for the age groups 1012, 1314, 1516 and 1719, with summary reporting for the age groups 1014 and 1519. At a minimum, statistically reliable information should be obtained and reported for the age groups 1014 and 1519.      W% Tobacco consumption levels Tobacco consumption levels may be reported in various ways. For daily cigarette smokers, the analysis should include the mean number of cigarettes smoked per day by smokers as well as the distribution of smokers according to some classification of cigarettes smoked per day. This could be, for example, <4, 514, 1524, 25 and over; or 17, 812, 1317, 1822, 2327, and 28 or more. These categorizations are suggested to-1n/n/ avoid a 'digit preference' among smokers who often tend to report consumption in numbers which are multiples of 5 or 10, i.e. 5, 10, 15 or 20 cigarettes per day. If the number of smokers is small, the categories may be collapsed in a way which is judged most appropriate to overcome this problem.  W.  Using Survey Data to make Population Estimates In many countries, the data obtained from probability samples are used to estimate the number of people in the entire population who use tobacco. In order to make such population estimates, data (or projections) on the sociodemographic characteristics (usually age, sex, and ethnicity) of the population is required. A set of "weighting factors" is applied to the survey sample data to permit the calculation of population estimates. The weighting factors are used to adjust the survey data so that they accurately reflect the characteristics of the general population. When this technique is applied, the data for one individual in the sample could actually represent that of thousands of people in the population. In order to conduct such a procedure, an appropriate statistical text book or a sampling statistician should be consulted. The calculation of confidence intervals and tests of statistical inference require specialized software programmes when weighted data are used. Countries that use unweighted data should report estimates separately for males and females, because the proportion of each sex in the sample is usually different from that in the overall population. Countries that use weighted data can report estimates for subgroups, and for the total population.  X$  ADDITIONAL QUESTIONS ON TOBACCO USE, KNOWLEDGE AND  X,& ATTITUDES Where resources are available, it can be extremely useful to ask additional questions on tobacco use, knowledge and opinions. C*-++Ԍ X  Tobacco use  X] 1)XAge at which first experimented with tobacco and age at which daily tobacco use began. This question should be given priority for inclusion in surveys dealing with tobacco use initiation;%" 2)XBrand of cigarette most commonly used, information on tar levels of cigarettes regularly smoked, and whether they smoke filtertipped or nonfiltered cigarettes;%" 3)XInterest in quitting. This could include questions such as: During the past 12 months, have you quit smoking for one day or longer because you were trying to stop smoking? Would you like to completely stop smoking cigarettes?%" 4)XWhether any health professional has given advice to stop smoking (doctor, dentist, nurse, pharmacist);%" 5)XIndicators of nicotine dependence such as time to first cigarette after waking. The following categories are recommended to classify responses: < 5 minutes, > 5 but < 30 minutes, 3060 minutes, > 60 minutes;%" 6)XExposure to environmental tobacco smoke at home, the worksite and public places. This could include the number of hours exposed and whether a member of the family smokes at home.%"  X^  Health Knowledge Understanding the level of public appreciation of the magnitude of the health risks of tobacco use can greatly assist tobacco control efforts. Information is a powerful influence on behaviour, thus strategies are needed which give people sufficient information to make informed decisions about their health. Finding out exactly what people know is necessary to determine further steps for package warnings, health education campaigns, etc. Thus, an important consideration in surveying health knowledge is whether or not smokers have made their decision to do so with sufficient knowledge of the potential risks involved: F* -++ 1. Recall of the diseases caused Simply knowing that tobacco is "bad", or responding positively when asked if tobacco use causes a particular disease does not show the extent of understanding critically important health information. This is one instance where an open ended question, such as the following, is recommended: "To the best of your knowledge, what, if any, are the health hazards related to smoking?" When this question was asked in a 1990 Canadian survey, only 20% of respondents identified heart disease as a potential health hazard.  X 2. Awareness of the probability of contracting such diseases Respondents may recall  XE that smoking causes lung cancer but believe there is only minimal risk of their getting the disease. 3. Awareness of the prognosis The following question would determine awareness of the seriousness of tobaccoinduced disease: "To the best of your knowledge, what percentage of lung cancer cases result in death?" According to the 1990 Canadian survey, onethird of respondents reported that they did not know, and only 14% were able to give the correct answer (i.e. more than 80%). The vast majority underestimated the risk. 4. Awareness of the benefits of quitting If an individual believes that quitting smoking will not reduce the risk of a heart attack ("I've smoked for 30 years so the harm is already done"), they cannot make informed decisions.  X^       Public Opinion Public Opinion Surveys (POS) can support and assist government policy concerning tobacco control legislation. In most cases, there is widespread support for comprehensive tobacco control measures, particularly those which prevent children from starting smoking. Often, it will be found that smokers also support tobacco control activities, as they usually do not want their children to smoke. The results of POS can also be a very powerful means of countering tobacco company arguments. F*!-++ԌAlthough POS can be combined with surveys on prevalence or health knowledge, they are usually undertaken as separate surveys, and are best related to current political possibilities for tobacco control action. For example, if a government is proposing a ban on tobacco promotion, it is useful to design a questionnaire focusing on this issue. POS can also help governments decide upon alternatives, such as whether to raise taxes on tobacco or on another product. A major advantage of POS is that they can be conducted in a short period of time. POS can be conducted by governments, but are usually conducted by university research departments, health institutes, or market research organisations. Because POS may serve as a powerful support for tobacco control, the survey design must be of high standard, with the questions free from bias. In order to increase the utility of the findings, data should be sought and analyzed according to such subgroups as age, sex, and perhaps political affiliation (where applicable), as well as smoker/nonsmoker status. The following are examples of commonly surveyed topics:  Smokefree areas  Ban on sales to minors  Ban on street vendors  Ban on free samples  Tobacco tax  Use of tobacco tax to fund health promotion and sports/arts sponsorship  Bans on tobacco advertising/promotion  Health warnings  Examples of public opinion surveys which were conducted in Canada may be found in Appendix VIII.3. C*"-++Ԍ X  Workplace Surveys Surveys which involve the entire workforce in the decision whether or not to implement smoking bans (or restrictions) are a key component to their successful implementation, providing that the final decision reflects, as closely as possible, the wishes of the workforce. A survey of all employees prior to the implementation of a smokefree workplace can identify support for this type of policy, assess current smoking patterns, areas of concern, and desire for assistance with cessation of smoking. Once a nonsmoking policy is implemented, surveys can then monitor compliance. An example of a workplace survey may be found in Appendix VIII.4.  X  ADDITIONAL PRECAUTIONS WHEN CONDUCTING SURVEYS When selecting subjects for prevalence surveys, every attempt should be made to eliminate selfselection, since it is generally accepted that selfselected individuals may not be representative of the population. Therefore, any prevalence data obtained using a sampling process where selfselection was involved may provide a biased estimate. It is difficult to obtain an idea about the magnitude, or even direction of the bias, if it exists. In any interview situation, it is important to minimize the number of nonresponses. Some individuals may not be available when contacted for an interview, or may refuse to take part. It is best to avoid the temptation to select additional individuals as a part of the sample to make up for the lost sample size. Since it is generally accepted that respondents are different from nonrespondents, a high nonresponse rate would substantially bias the estimated prevalence. For example, in a housetohouse survey, the subgroup that is often most difficult to interview are working males. Their smoking patterns may be very different to other sectors of the population. In this case, it is advisable to commit extra resources in order to try to interview the nonrespondents, through extra efforts. C*#-++ԌWhenever feasible, facetoface interviews should be conducted in private, using responses obtained directly from the interviewee. If other individuals are within hearing distance of the interview, the interviewees may not give accurate answers. As discussed earlier, in some cultures, women may not be comfortable talking about their tobacco habits in front of others. In other cases, young people may not give accurate responses concerning their tobacco habits in front of parents or elders.  X  KEY ISSUES: PREVALENCE SURVEYS  XE  * XAccurate data on the prevalence of tobacco use in the total population, among men and women, and among specific subgroups are the most important and useful measures of the magnitude of the tobacco epidemic in a given population."  X  * XProbability sampling is the only scientifically valid procedure for estimating prevalence in a population."  X  * XIt is extremely important to avoid selfselection in the sample, and to minimise the level of nonresponse."  X  * XPrevalence surveys should be repeated at least every five years, and if possible annually, in order to monitor trends in tobacco use behaviour." $-++  X  6&6&StandardHPLAS4.PRSx  &6&6StandardHPLAS4.PRSx   %      Table VIII.1 Strengths and weaknesses of various data collection approaches T ddx]  ddx%     T "    " z N 6Developing Countriesz " &qDeveloped Countries   Z   Type of survey and mode  -Strengths IWeaknesses  !gStrengths  X (*Weaknesses ăZ C z Interview/Household . ܩ Questions are easily explained Higher response rate  X\ ԩ Most expensivea Less privacy may lead to under reporting ܩ Questions are easily understood Higher response rate  X\ ԩ Most expensivea Less privacy may lead to under reportingC   Interview/Telephone ܩ Cheaper than household Questions are easily explainedC  ܩ Few households might have telephones Difficult to obtain a representative sample Less privacy may lead to under reporting  ܩ Cheaper than household High telephone coverage minimizes bias ܩ Less privacy may lead to under reporting   SelfAdministered Questionnaire/ HouseholdX ܩ Questions can be explained and responses can be clarified Higher response rate* ܩ Much more expensive than mail High illiteracy rate may require assistance with the questionnaire (increasing the cost and reducing privacy) ܩ Privacy Responses can be clarified Higher response rate than mail ܩ Much more expensive than mail .   SelfAdministered Questionnaire/ Mailo ܩ Privacy Relatively cheap Avoids interviewer biasX ܩ High illiteracy rate Skip patterns may be difficult to comprehend Very low response rateX ܩ Privacy Relatively cheap, allowing larger samples Avoids interviewer bias* ܩ Low response rate in some cultures Responses cannot be clarified and skip patterns may be hard to understand.   &6&6StandardHPLAS4.PRSx  &6&6Standard/rJet 4E & %%     #Xw P7 XP#  Y a Among all survey options%  X  Table VIII.2 aa Required sample size for estimating smoking prevalence (P) with specified precision (d) using a simple random sample and a 95% confidence interval. 3a ddx%     ddx& $.   c $  P d   5%  *10% 715% ~C20% qP25%  XL d]30% Ã  XL  W!i40% $J%v50%  (=)60%  s,0-70%  N0 180%c      Y  + 1%  + 2% + 3% + 4% + 5% + 10%   1825 456 203 114 73 18  3457 864 384 216 138 35  4898 1225 544 306 196 49  6147 1537 683 384 246 61  7203 1801 800 450 288 72  8067 2017 896 504 323 81  9220 2305 1024 576 369 92  9604 2401 1067 600 384 96   9220 2305 1024 576 369 92   8067 2017 896 504 323 81   6147 1537 683 384 246 61      Y  d is the precision required (in percentage points) on either side of the prevalence estimate P . It is recommended that d be no greater than + 5%, and preferably smaller.  YP  For example , if the anticipated prevalence of smoking in a population (or population subgroup) is 20%, and we require that the true prevalence be estimated within 5 percentage points (with 95% confidence), then we would require a sample size of 246. Normally this would be only for a specific subgroup such as males aged 1524, and hence the total sample size for a general population survey would have to be much larger to permit sufficiently reliable estimates for each agesexcategory of the population. As a minimum, a sample size of about 2000 would be required (1000 men, 1000 women) to give reasonably reliable estimates for broad age groups and for men and women separately. For sample sizes based on more complex sampling designs and other confidence intervals, see Lwanga and Lemeshow, 1991. &```  X  &6&6Standard/rJet 4E &6&6&Standard/rJet 4E &%'  # \  P6B& P# #Xi\  P6ƒXP#      Appendix VIII.1 Case studies of prevalence surveys  X a1 Case study I: TIFR Studies   The Tata Institute of Fundamental Research has conducted studies in several rural populations of India to get a reliable estimate of the prevalence of tobacco use and oral precancerous lesions. A district was chosen as the basic geographical unit, and five districts (each with a population ranging from over a million to four million) were selected. A list of villages was prepared from the District Census Handbook published by the census department. Villages that had a population of over 2500 or less than 500 were excluded from the list. Utilizing information on the population distribution of the villages and age distribution of the population, the number of villages that would approximately provide the sample size of at least 10 000 individuals aged 15 years and over, was calculated. These villages were selected by the method of simple random sampling.   A team consisting of a dentist and interviewers went housetohouse in the selected villages and interviewed every individual over the age of 15 years. They visited each household at least twice in an attempt to minimize the nonresponse rate. After completing all villages they made one more round of all the villages in an attempt to interview the remaining persons.   This procedure provided reliable estimates of tobacco use prevalence in parts of rural India. For example, it established for the first time that chewing generally meant chewing tobacco; the number of chewers who chewed betelquid or areca nut without tobacco was negligible.   The limitation of this approach was that it was expensive and time consuming, particularly because the random sampling selected a large number of villages in interior areas that were not easily accessible. It was possible, however, to carry out the survey because the main objective was to investigate the epidemiology of oral precancerous lesions in this population. Z.'n/n/n/Ԍ  X   a,Case Study II: Prevalence in Malaysia ă   The Dental Public Health Division of the Ministry of Health in Malaysia carries out sample surveys routinely to estimate the extent of dental problems in the country. Following a suggestion from the Dental Faculty of the University of Malaya, this Division decided to undertake a survey of oral mucosal lesions in Malaysia. To get a representative sample of the entire country, they adopted a stratified two stage sampling procedure.   The smallest unit available from the census office was an "enumeration block" consisting of 80 to 120 households. This was used as the sampling unit. The country was divided in 4 strata: general metropolitan areas, large urban areas, small urban areas, and rural areas. Within each stratum, enumeration blocks were selected by random sampling, and within each enumeration block, a systematic sample (such as every 10th house) of living quarters was selected (with a random start). With a target sample size of 11 000 to 12 000, all adults aged 25 years and above in selected living quarters were interviewed and examined. Institutions such as hostels, hospitals, prisons, boarding houses, and military barracks were excluded.   It would have been quite difficult for an individual investigator to undertake such a survey because of the resources and expense involved, but this was less problematic for the Ministry of Health. With technical expertise from a local University, it was possible to modify a routine survey project in order to get reliable estimates of tobacco use prevalence.  Xr#  a.Case Study III: Hospital patients ă   A Danish professor, teaching dental students in India, wanted to get some idea of the prevalence of oral lesions and their association with various form of tobacco use in the country. With the help of local faculty, he interviewed and examined consecutively the first 10 000 dental outpatients treated at the dental college of each of the cities of Lucknow, Bombay and Bangalore, and 5 000 in Trivandrum. Z.(n/n/n/Ԍ  Although the authors were aware that the sample was biased, particularly with regard to age and gender, they maintained that since private dental care is rare and costly, the socioeconomic status of the patient group was probably representative of all but the very wealthiest.   The survey revealed a great variety of tobacco use. There were seven different tobacco practices (smoking cigarettes, bidi, chilum, hookah, and chewing of tobacco and pan with and without tobacco). A total of 38 combinations of these various forms of tobacco use were listed. It became obvious that for any survey of tobacco use in India, one must be careful about how tobacco use is defined and to enumerate combinations of different kinds of tobacco use.  X\  a- Case Study IV: Bombay Police Study ă   As part of a study on leukoplakia, a survey on tobacco use among the police in Bombay was conducted in 1959. Policemen were selected because they formed a fairly homogeneous group almost all of them were originally from the same district.   Because of his position as honorary dental surgeon to the police, the principal investigator was able to obtain the cooperation of the highlevel officers for this study. The policemen were ordered to report to the Police Hospital during appointed dates and times in convenient batches. In this way, he was able to interview and examine (for oral precancerous lesions) 4734 policemen in a relatively short period of time.   Despite the fact that the results were not generalisable, the surprising extent of tobacco use (over 85% aged 26 years or over chewed or smoked tobacco) was of interest. The group was followedup after 5 and 10 years. This was the first longterm followup study of tobacco use in a cohort in India. These results were later corroborated in populationbased followup studies of large cohorts in rural areas in India. ()n/n/n/  X  Appendix VIII.2 An example of methods of sampling for a national, prevalence survey with reliable data  X for both rural and urban areas Sampling for "two" national (one urban, one rural) surveys 1.X  Obtain a list of districts. Suppose that about 80% of the districts are rural and 20% urban.#  2.X  Separate into:#    i.aaRural districts rural stratum#a   ii.aaUrban districts urban stratum#a 3.X  For each stratum: select randomly 20 districts#    i.aaRural: select 20 districts#a   ii.aaUrban: select 20 districts#a It is better to select districts based on the principle that the probability of being selected is proportional to size (number of persons in the district). 4.X  For each rural district, randomly select 10 communes#  X  For each urban district, randomly select 12 precincts#  5.X  For each selected commune, work out the number of families (households), as shown below.#  For each precinct, select randomly 45 wards. 6.X  Assume three persons aged 15+ in each household.#  7.X  Total number of persons to be interviewed:#    i.aaRural 4 000#a   ii.aaUrban 4 000#a   This means 4000/3 = 1 400 households in each of the rural and urban areas. 8.X  For rural areas: 1 400 households/20 districts = 70 households per district or about 7 households per commune.#  9.X  For urban areas: for 70 households per district or 6 households per precinct, select randomly 12 households per ward.#  10.  Visit the household selected and interview all persons living there aged 15 or above. Z.*n/n/n/Ԍ X  6&6&Standard/rJet 4E &&6&6Standard/rJet 4E & '+ &6&6Standard/rJet 4E &&6&6Standard/rJet 4E & ++ #Xw P7 XP#  \    ^ =  ==  &6&6Standard/rJet 4E &6&6&Standard/rJet 4E &,^\++ Appendix VIII.3 Sample questions for public opinion surveys, Canada, 19791987       Y   aa "  )ii18@qqGO Percentage in Favour   aa "  )ii18@qqGO2   aa "  )ii18@qqGO!!Vyy^e)")"m$$t$$t$$t$$t$$t$$t$$t$$t$$t$$t$$t aa "  )ii18@qqGOTotal!!Vyy^NonSmokers)")"mSmokers   aa "  )ii18@ qqG O % yy^e %)")"m %  XI  1979 Is there a need for control by federal government legislation of: a)X  Reporting of constituents on the packaged%  X  (for example, tar, nicotine,d%  X  carbon monoxide)?  )ii18@ qqGO80!!Vyy^ e82)")"m 79d%  b)X  Health warnings on cigarette packagesd%    and/or advertising?  )ii18@ qqGO80!!Vyy^ e83)")"m 77 c)X  Tar and nicotine levels of cigarettes? @ qqGO72!!Vyy^ e74)")"m 71d%  d)X  Levels of other substances in cigarettes?@ qqGO66!!Vyy^ e68)")"m 65d%  e)X  Advertising of cigarettes?ii18@ qqGO65!!Vyy^ e69)")"m 62d%  f)X  Promotion of cigarettes?ii18@ qqGO63!!Vyy^ e68)")"m 59d%   X  1981 Would you support: a)X  Legislation to control the tar andd%  X  nicotine contents of cigarettes?8@ qqGO84!!Vyy^ e86)")"m 82d%  b)X  Legislation to control the salesd%    of cigarettes or the number of cigarettes   sold yearly? "  )ii18@ qqGO41!!Vyy^ e51)")"m 31  X!  1982  1.X  Do you think there should be nonsmokingd%    areas on buses, trains and airplanes?@ qqGO93!!Vyy^ e95)")"m 90 2.X  Do you think the amount of advertising ofd%    cigarettes should be reduced to a lower   level or eliminated altogether?8@ qqGO64!!Vyy^ e69)")"m 56 3.X  Would you agree that:d%  a)X  The government should ban all advertisingd%    for cigarettes "  )ii18@ qqGO42!!Vyy^ e)")"m b)X  The government should ban cigarettes ind%    any public place?  )ii18@ qqGO54!!Vyy^ e)")"m U/+=000  X  =Appendix VIII.3   aa "  )ii18@qqGPercentage in Favour   aa "  )ii18@qqG2   aa "  )ii18@qqGTotalONonSmokersyy^eSmokers   aa "  )ii1 8 @qqG %O %yy^e %    Xv  1985 Is the measure in the Federal Budget to increase the price of a 25 pack of cigarettes by 25 cents immediately good for the country?aa "  )ii18 @qqG75O!!V yy^e  X  1986  Would you agree that: a)X  The government should ban alld%  X  advertising for cigarettesii18 @qqG63O!!V yy^e d%  b)X  The government should ban cigarettesd%  X  in any public place?  )ii18 @qqG60O!!V yy^e d%  Do you think smoking in the workplace should be: a)X  generally allowed?  )ii18 @qqG18O!!V 9yy^e 31d%  b)X  in designated areas?  )ii18 @qqG57O!!V 58e 54d%  c)X  not allowed? "  )ii18 @qqG22O!!V 30e 12d%  Do you think smoking on airplanes should be a)X  generally allowed?  )ii18 @ qqG 6O!!V 3yy^e 10d%  b)X  in designated areas?  )ii18 @qqG51O!!V 43e 63d%  c)X  not allowed? "  )ii18 @qqG41O!!V 52e 24d%   X  1987 Would you say that you favour or oppose a law banning smoking in the workplace except in private offices and smoking rooms? "  )ii18 @qqG70O!!V yy^e Do you approve or disapprove of the proposed law to ban advertising of cigarettes and all other tobacco products* "  )ii18May @qqG68O!!V 74yy^e 56   aa "  )ii18November qqG67O!!V 74yy^e 54 Would you support or oppose a total ban by the government on the sale of tobacco products?@ qqG38O!!V 46yy^e 24 * (persons with no opinion excluded)   aa "  ) @ qqGk.,=000  X # Xi\  P6ƒXP#   Appendix VIII.4 7SMOKEFREE WORKPLACE Q7SAMPLE QUESTIONNAIRE 1.  Which of these phrases best describes your own view about smoking at your place of work. (Tick one). Smoking should not be allowedii18@qqGO!!V_ There should be separate areas where smoking is permittedO!!V_ Smoking should be allowed in all areas8@qqGO!!V_ Don't knowaa "  )ii18@qqGO!!V_ 2.  What do you prefer in the areas where people work together? (Tick first and second preference).aa "  )ii18   aa "  )ii18@qqGO1st 2nd No restrictions on smoking  )ii18@qqGO_!!V_ Separate smoking and non-smoking working areasqqGO_!!V_ No smoking except at break times and designated areasqqGO_!!V_ Total ban on smoking in working areas8@qqGO_!!V_ Other (please specify).............................. .................................................................. 3.  What do you prefer in the following areas? (Tick one in each line).  aa Total  )Separate8 Smoking atqqGONo  aa Ban "  )Areasii18 certain timeqqGORestriction Toilet  aa "_  )ii1_8@_qqGO!!V_ Canteenaa "_  )ii1_8@_qqGO!!V_ Tea loungeaa "_  )ii1_8@_qqGO!!V_ Sports areaaa "_  )ii1_8@_qqGO!!V_ Lifts  aa "_  )ii1_8@_qqGO!!V_ Corridorsaa "_  )ii1_8@_qqGO!!V_ Other........aa "_  )ii1_8@_qqGO!!V_ 4.  At meetings, which do you prefer? (Tick one).  Total ban "  )ii18@qqG_  Smoking breaks  )ii18@qqG_  Majority vote  )ii18@qqG_  No smoking unless all agree8@qqG_  No restriction  )ii18@qqG_+-000  X  Appendix VIII.4 5.  Which of the following describes your working area best? (Tick one).  Private office "  )ii18@qqG_  Shared office "  )ii18@qqG_  Open-plan office  )ii18@qqG_  Shop floor "  )ii18@qqG_  Other (please specify)ii18@qqGO!!Vyy^e 6.  Which of the following describes you best? (Tick one).  I am a smoker who wants to give up@qqG_  I am a smoker who doesn't want to give upqqG_  I am an ex-smoker  )ii18@qqG_  I am a non-smoker  )ii18@qqG_ 7.  Is smoking permitted in your work area?  Yesaa_ "  )Noii1_ 8.  Are you bothered by tobacco smoke at work?  Yesaa_ "  )Noii1_ 9.  If you are bothered by smoke at work how does it affect you? (Tick any that apply)  Worries about long-term effect on health@qqG_  Eye irritation "  )ii18@qqG_  Headache "  )ii18@qqG_  Coughing "  )ii18@qqG_  Stuffy or runny noseii18@qqG_  Breathing difficulty  )ii18@qqG_  Loss of concentrationii18@qqG_  Clothes and hair smellii18@qqG_  Other (please specify)ii18@qqGO!!Vyy^e 10.  Have you ever had to move away from where you were working because of other people's smoke?  Frequently "  )ii18@qqG_  Occasionally "  )ii18@qqG_  Neveraa "  )ii18@qqG_ +.000  X  Appendix VIII.4 FOR SMOKERS ONLY: NON-SMOKERS PLEASE GO TO END OF QUESTIONNAIRE 11.  Do you smoke in your work area?  Yesaa "  )ii18@_  Noaa "  )ii18@_ 12.  If you could not smoke in your work area, would you smoke less, or stop?  Smoke less "  )ii18@_  Stopaa "  )ii18@_   No change "  )ii18@_ 13.  If you could not smoke at all in your working area, what would it be like for you? (Tick one).  Very easy "  )ii18@_  Easyaa "  )ii18@_  Difficult "  )ii18@_  Very difficult  )ii18@_ 14.  Would you use help to give up smoking if it were offered at work?  Yesaa "  )ii18@_  Noaa "  )ii18@_ Thank you for completing the questionnaire. Please add any comments here:ii18@qqGO!!Vyy^  aa "  )ii18@qqGO!!Vyy^  aa "  )ii18@qqGO!!Vyy^ Please return it to: "  )ii18@qqG By (date):aa "  )ii18@qqG