SlideShare a Scribd company logo
1 of 9
Download to read offline
HEALTH POLICY AND PLANNING; 17(3): 322–330                                                      © Oxford University Press 2002




How to do (or not to do) . . .

Methods for pre-testing and piloting survey questions:
illustrations from the KENQOL survey of health-related quality
of life
A BOWDEN,1 JA FOX-RUSHBY,1 L NYANDIEKA2 AND J WANJAU2
1Health Policy Unit, London School of Hygiene and Tropical Medicine, UK, and 2Centre for Public Health Research,

Kenya Medical Research Institute, Nairobi, Kenya

   Part of the assertion that any survey researcher can make about the validity of their results needs to contain
   an analysis of questions and their responses from the respondent’s viewpoint. Claims concerning the valid-
   ity, reliability and sensitivity of health-related quality of life measures tend to be based on the quantitative
   approach of psychometrics, which fails to identify when respondents: misinterpret questions; do not recall
   the information requested; or give answers that present themselves in a better or worse light. The paper
   presents some approaches to pre-testing and piloting survey questionnaires to check the interpretation of
   survey questions, using illustrations from the KENQOL project. The paper describes: how the intended
   referential and connotative meaning of each question was established; the criteria to judge the appropriate-
   ness of each question; the methods used to make those judgements; and the process of reviewing questions
   based on findings. The role of piloting is highlighted, and further reading is suggested for readers wishing
   to develop a model for their own investigation.

Key words: survey, pre-testing, piloting, Kenya, health-related quality of life




Introduction                                                        We were in the process of developing a survey tool to
                                                                    measure the self-perceived and locally defined ‘health’ of
Questionnaire-based surveys are a very common approach to           people in Makueni District (Kenya), and reviewed practices
data collection used by researchers and the interpretation of       in the development and translation of such measures as part
findings may be used to inform policy or public debate. It is        of this process. We noticed that claims concerning the valid-
not uncommon to find reports in the media saying ‘a recent           ity, reliability and sensitivity of health-related quality of life
survey showed that X% of people think . . . .’ But do they?         (HRQL) measures tended to be based on the quantitative
How much trust can we place in the findings of surveys?              approach of psychometrics. Such quantitative techniques are
Answering questions on a survey requires several thought            unlikely to identify when respondents misinterpret ques-
processes: respondents have to interpret the question,              tions, fail to recall the information requested or give answers
retrieve the necessary information and then decide what             that present themselves in a better light. Indeed the interpre-
answer to give – and all this before the researcher gives their     tation of questionnaire items has often been overlooked
own interpretation to the responses.                                (or at least not reported) in the development and trans-
                                                                    lation of HRQL measures (Fox-Rushby and Parker 1995;
Part of the claim that any survey researcher can make about         Bowden and Fox-Rushby 2000). It is also the case that well-
the validity of their results needs to contain an analysis of the   known texts in survey research often fail to address issues
questions from the respondent’s viewpoint. This is a salient        of how to pre-test survey questions in sufficient detail
starting point either for those developing their own question-      (Moser and Kalton 1992; Fowler 1993; Grosh and Glewwe
naire or for those using an existing questionnaire in a new         2000).
context (either in a new population or new language). Failing
to investigate the interpretation of questionnaire items may        The aim of this paper is to present a variety of examples to
result in misinterpretations (by respondents and researchers),      help researchers improve the validity,1 reliability2 and sensi-
falsified answers, missing responses (to the questionnaire as a      tivity3 of their survey instruments prior to undertaking a
whole or to particular items), and possibly an offended             household survey. The paper first provides a brief context to
respondent who chases away the interviewer and encourages           our research before proceeding to the range of methods used
others to refuse interviews.                                        to pre-test and pilot the survey instrument. The discussion
How to do (or not to do) . . .                                               323

section focuses on the generalizability of our approach to        Pre-testing the survey questions
testing questions for surveys and introduces the reader to a
                                                                  Pre-testing is the main chance for researchers to gauge the
range of literature for further reading.
                                                                  meaning attributed to survey questions ‘before it is too late’,
                                                                  i.e. before a substantial investment is made in the wrong
Context of research                                               questions or in questions where the researcher cannot be
                                                                  sure about what is being asked. It is therefore imperative that
The overall aim of our research is to develop a culturally
                                                                  the researchers undertake these interviews, rather than
relevant generic measure of ‘health’ to measure the impact of
                                                                  assigning the task to field assistants, because they know the
interventions designed to reduce disease and/or improve
                                                                  aim of the research. The pre-testing involved a number of
health in Kenya. We have been working amongst the Kamba
                                                                  steps including:
of Makueni District in Eastern Kenya and the Maragoli of
Vihiga District in Western Kenya. The conceptualization and
                                                                  • establishing the intended referential and connotative
operationalization of ‘health’ emerged from extensive
                                                                    meaning of each question;
periods of participant observation by anthropologists, in
                                                                  • agreeing a set of criteria to judge the appropriateness of
addition to qualitative research by others (see Fox-Rushby et
                                                                    survey questions;
al. 1997; Fox-Rushby, for the KENQOL Group, 2000). The
                                                                  • selecting the methods for judging appropriateness and
definition of health adopted focused on the subjective per-
                                                                    undertaking research;
ceptions of the positive and negative aspects of health that
                                                                  • reviewing questions for inclusion, revising (the question or
comprised ‘contentment’, ‘cleanliness’, corporeal capacity’,
                                                                    intended meaning) or exclusion.
‘co-operation’ and ‘completeness’ of an intending individual
within the intentional worlds of Makueni and Vihiga (Fox-
Rushby, for the KENQOL Group, 2000).                              Step 1: Establishing the intended referential and
                                                                  connotative meaning of each question
The KENQOL survey questionnaire was developed to
                                                                  It is important to ascertain whether respondents are inter-
access those aspects of health that a member of the com-
                                                                  preting the question as intended to ensure appropriate
munity could respond to in public4 to an interviewer who was
                                                                  interpretations are placed on the data. Therefore we pre-
not necessarily known to them personally. This paper focuses
                                                                  pared descriptions of the intended referential and connota-
on our work amongst the Kamba and reflects the develop-
                                                                  tive meaning for each of the survey questions drawn up.
ment of the household survey in the Kamba language.5 The
                                                                  There were several advantages to this: it served as an account-
aim of this part of our research was to investigate: how
                                                                  able approach to assessing content validity; it facilitated
respondents understood the KENQOL survey questions;
                                                                  clarification of our own intentions in asking questions; it
the types of information retrieved and over what time
                                                                  ensured we assessed the same hypotheses as there was less
period; the types of socially acceptable answers that might
                                                                  variation in our own interpretation of the meaning of the
be given and why; in addition to evaluating whether they
                                                                  original questions (either over time or between researchers);
performed the task the way the researcher expected. We
                                                                  and it was expected to be helpful in the future assessment of
tested a total of 150 potential health measurement survey
                                                                  the conceptual equivalence of the Kamba language version
questions.
                                                                  with other language versions (see Herdman et al. 1997, 1998).
                                                                  Box 1 provides an example of the type of description we
The field site was located in Makueni District in Eastern
                                                                  attached to a question.
Kenya. It is approximately halfway between Nairobi and
Mombasa, just north of the main trunk road. The Kamba
population (part of the Central Bantu group) is the majority
ethnic group in this region (99%) and traditionally they are
pastoralists. Currently 50% of the population are aged
0–14 years but infant mortality rates are high, at 96 per 1000
live births in 1995 (Kenya Government 2000). Rainfall is          Box 1.   Sample meaning for a KENQOL survey question
scarce in the area, with an annual precipitation of less than
500 mm and this, combined with high daytime temperatures,
leads to a hot, dry environment. Agriculture is unreliable and       Do you feel that there are some people who could sabo-
food deficits are often experienced. Health facilities are rela-      tage the progress of people in your neighbourhood at the
tively scarce, although they can be reached by matatu from           moment?
the main Nairobi/Mombasa highway; the walk to the road can
take up to 2 hours.                                                  We want to find out an individual’s current views about
                                                                     whether they feel that their ability to improve them-
A day’s bus journey is required to reach hospital facilities in      selves and their lives (collectively as a community)
Nairobi. The major diseases are malaria, respiratory tract           may be adversely affected by people within the com-
infections, diarrhoeal diseases, skin ulcers, intestinal worms       munity in some way. It is possible this may include a
and urinary tract infections (Kenya Government 2000).                range of ideas such as violence and witchcraft and we
Incomes are low compared with many other areas in Kenya,             leave it up to the respondent to decide if they are being
with many people living below the poverty line and involved          threatened.
in a daily struggle for survival and development.
324                                                   How to do (or not to do) . . .

Table 1. Criteria selected for judging appropriateness of survey questions and investigative method employed

1 Criteria                                                                              2 Method

                                                                                        Expanded     Targeted    Group       Expert
                                                                                        interviews   methods     methods     reviews

 1    No negative questions, and therefore no double negative answers                   √                                    √
 2    There are not two questions in one                                                √                                    √
 3    Level of language is not too high/old fashioned/unusual in any sense              √                                    √
 4    The question is simple and grammatically correct                                                                       √
 5    The question is free from jargon                                                                                       √
 6    Singular and plural ‘you’ is clear                                                √                        √           √
 7    The same idea is not contained in another question                                √                        √
 8    The question is sensitive to measure change over time                             √            √
 9    Captures current views                                                            √                        √
10    Reflects local issues                                                              √                        √
11    Meaning and interpretation of question is clear                                   √            √           √
12    The question makes sense to everyone                                              √                        √
13    A ‘yes’ is clear and unambiguous in meaning                                       √                                    √
14    A ‘no’ is clear and unambiguous in meaning                                        √                                    √
15    Time period is clear                                                              √                        √           √
16    Conception(s) of health it represents is specified                                              √                       √
17    How the question relates to: time of year; head of household; individual/
      homestead/wider family/community                                                  √                        √           √
18    Understanding of how it relates to severity of HRQL                               √            √




Step 2: Criteria to judge the appropriateness of survey                  organize the expanded interviews we divided the set of survey
questions                                                                questions into groups of between three and five questions. In
                                                                         each interview6 the respondent was first asked the KENQOL
Having developed the survey question and agreed on the
                                                                         survey question and asked to respond either ‘yes’, ‘no’ or
intended meaning, we developed a series of criteria for evalu-
                                                                         ‘don’t know’ as they chose. The probes used were divided into
ating the appropriateness of the survey questions. This facili-
                                                                         three types and used in separate interviews:
tated an agreed and standardized way of making decisions
about which questions to include and enabled different
                                                                         (1) Comprehension of ‘key phrases’ within the question. We
researchers to bring their own judgements together in a
                                                                             asked respondents to talk about the kinds of things they
comparable way. Column 1 of Table 1 outlines the range of
                                                                             thought about when answering the question. Examples of
criteria we adopted. It can be seen that the issues focus largely
                                                                             key phrases included the time frame the respondent
on the style of language, attributed meaning, clarity of the
                                                                             considered and who respondents included when the
question and consistency of interpretations.
                                                                             word ‘friends’ or ‘neighbours’ was used (see Box 2 for an
                                                                             example).
                                                                         (2) Applicability of the questions: to people of different ages,
Step 3: Methods for judging appropriateness of survey
                                                                             gender and socioeconomic groups (see Box 3 for an
questions
There are a wide variety of techniques available to
researchers wanting to test survey questions. We adopted
four principal types (described in Forsyth and Lessler 1991):            Box 2.   Comprehension of key phrases in the question
expanded interview techniques, targeted methods, group
methods and expert evaluation. Others approaches are
addressed in the discussion. Table 1 (column 2) indicates                   Do you feel safe at the moment?
which technique(s) addresses which type of issue and this
section outlines each method. Of these techniques, the                      The question intended to access concerns about per-
expanded interviews took most of our time.                                  sonal safety and safety of belongings. The probe ques-
                                                                            tions asked respondents to give examples for why
                                                                            someone (a) will feel safe and (b) may not feel safe. It
Expanded interviews
                                                                            emerged that the question not only addressed notions
Expanded interviews begin by one or more of the intended                    of personal safety and safety of belongings, but also
survey questions being asked and answered by the respon-                    having enough food, being properly taken care of and
dent, followed by a set of probe questions to investigate the               having money for school fees.
thought processes involved in answering the question(s). To
How to do (or not to do) . . .                                               325

Box 3.   Applicability of questions for respondents of different ages


   We asked respondents who in the family can and cannot be asked the question, ‘At the moment are you always worry-
   ing about getting food?’ The text below demonstrates some of our findings with respect to this question.

   Interview with an elderly man:

   Interviewer: Can children worry about getting food?

   Respondent: No, since it is their father’s fault they are fed.

   Interviewer: Could a young unmarried man answer this question?

   Respondent: A young unmarried man isn’t responsible for that. You will find that in most cases he comes at night and
   gets in his room and sleeps. Maybe he has passed to his friends’ home and had something to eat. The old mamas can’t
   worry since they are just looked after by their children they brought up.

   Interviewer: Is there anyone who might say no to this question?

   Respondent: Old and young people.




   example); by level of sensitivity of the question; and               you ‘clean’? – meaning in a hygienic and moral sense),
   social desirability of responses (see Box 4).                        but further probing highlights some potential problems.
   Box 4 has a particularly interesting response because it         (3) The meaning of a ‘yes’ or ‘no’ response needed to be clear
   also shows how sensitive the questioning on sensitivity              and as intended (see Box 5).
   needs to be. It might at first appear that there are no prob-
   lems with the question ‘Ivindani yii nutonya kwiyikalya          The respondents for the expanded interviews were volun-
   wi mutheu undu utonya kwenda?’ (At the moment, are               teers willing to expend more time than the data collection



Box 4.   Addressing the sensitivity of the question


  We asked respondents directly about the sensitivity of questions. We needed to know whether a question was so sensi-
  tive that people might purposefully provide the wrong answer. Alternatively they may provide the right answer, but this
  makes the respondent or their family uncomfortable in some way. The transcript below is a translation from an inter-
  view that addressed sensitivity.

  Interviewer: At the moment, are you ‘clean’ [hygienically and morally]?

  Respondent: Yes

  Interviewer: When we ask this question in the interview, we want to be sure that we are not embarrassing the person we
  are asking. Do you think this question might cause embarrassment if it is asked in an interview?

  Respondent: The question is good and cannot cause embarrassment in an interview.

  Interviewer: Do you think people will feel they can answer this question truthfully in an interview?

  Respondent: Not all people can answer truthfully. This is because if one feels unclean or has done wrong, he wouldn’t
  want to be known to have wronged.

  Interviewer: If someone answers ‘no’ to this question, how will it reflect on the family of that person?

  Respondent: One who answers no would be thought to be unclean, but not the whole family.
326                                             How to do (or not to do) . . .

Box 5.   Testing the meaning of a ‘yes’ or ‘no’ response


   By undertaking a relatively simple interview we could quickly establish whether a ‘yes’ or ‘no’ response could be inter-
   preted clearly and as intended. This particular question was dropped because it became clear that we could not be sure
   that a ‘no’ meant worse ‘health’.

   Interviewer: Do you get enough information about current events in Kenya?

   Respondent: No.

   Interviewer: If someone answers ‘no’ to this question, is he/she in a better or worse position than someone who answers
   ‘yes?

   Respondent: The one who answers ‘yes’ is in a better position.

   Interviewer: Why is this?

   Respondent: Information about current events in Kenya can assist people in this area by helping them develop, such as
   knowing different methods of farming, knowing about disease and outbreaks and ways of combating them, and also
   improvement of educational facilities.

   Interviewer: What other reasons are there that someone could answer ‘yes’?

   Respondent: Those who own radios and are literate. People can get access to this information from radios, newspapers,
   meetings and friends.

   Interviewer: What other reasons are there that someone could answer ‘no’?

   Respondent: Those who don’t have radios and don’t read newspapers. Everyone doesn’t want to have access to infor-
   mation about Kenya. Those who do are mostly interested in politics and current affairs. Those who aren’t are the old,
   illiterate and those not interested in politics.




itself in helping us learn about the questions. They were          poor; and those with harmonious relationships compared
selected purposively to include equal proportions of males         with those in disputes.
and females, as well as being roughly representative of the      • As the questionnaire was intended to be able to measure
different generations in the community. In total each ques-        change we asked eight people for one of two types of hypo-
tion was the focus of between four and eight interviews with       thetical considerations of the questionnaire. For example,
a range of people from the local population.                       members of a youth-based water and agriculture project
                                                                   were asked to imagine how they might have answered the
                                                                   questions before and after the project, and other key
Targeted methods
                                                                   informants were asked how they might have completed
Targeted methods focus on a particular component of the            the questionnaire during and after malaria. The task was
response process, to gain a detailed insight into the respon-      followed up with a discussion of the answers most likely to
dent’s comprehension and preparation of a response and             change and why.
to evaluate whether they performed the task the way the          • To address which conception of health people felt the
researcher intended. We used three types of targeted               question represented, as well as helping further investi-
methods:                                                           gations into the meaning and interpretation of the ques-
                                                                   tion, respondents were asked to sort the Kikamba
• People with known (and opposite) characteristics of inter-       questions into groups, either under the Kikamba words
  est were asked to complete specific questions to assess how       used to translate the English word ‘health’7 that each ques-
  likely it was that the questionnaire would be interpreted in     tion was most closely related to, or to devise their own
  the same way by people with particular characteristics. This     categories and group them. Each question was placed on a
  was important in being able to interpret any future differ-      separate card for ease of sorting. The task was followed up
  ences/similarities across groups. Therefore we contacted         with an interview about the meanings of the Kikamba
  (through the help of our field assistants): people with and       words used to translate the English word ‘health’ and
  without sickness/illness; the relatively wealthy and very        reasons for groupings.
How to do (or not to do) . . .                                              327

Box 6.   Group discussion on ‘people in your homestead’


    Many of the KENQOL questions ask respondents about people in their homestead,i for example,

    • Do people in your homestead listen to your advice/suggestions at the moment?
    • When you are sick, do people in your homestead help you in getting treatment?
    • At the moment is it easy for people in your homestead to find transport to go for treatment when they are sick?

    To check who people thought about when answering these questions, we provided several groups of respondents with
    an example family (see below) and initiated a discussion about who different people in that family would think about
    when they were asked specific questions.

    Example family: Headed by a married male (mzee) and female (mama), who had three sons. The first son is married
    to one wife with a child and they live just outside the compound. The second son is married with two wives (polygy-
    nous marriage) and both wives have children. The third son lives away in Nairobi.

    The discussion revealed that people would make reference to different people in their response. In the example above,
    the mzee would consider seven adults, whereas the first-born son thinks of his wife and his wife thinks only of her
    husband. To complicate matters further still, we discovered that a person could alter the group of people they thought
    about depending on the question that was being asked, even if each question used identical phrases for ‘people in your
    homestead’. We noted all the instances in which this occurred.

i The phrase musyi kwaku/kwenyu for homestead was used after finding that the word family was interpreted very
differently across people.




Group methods                                                    questions and issues 1–6 in Table 1. The local field assistants
                                                                 also reviewed the written Kikamba questions and gave their
Group methods also focus on a particular component of the
                                                                 views on issues 1–6 as well as 13–17. They were very helpful
response process, but a group of people are addressed, rather
                                                                 in their criticisms and suggestions, and once we had been able
than an individual, to encourage greater discussion around
                                                                 to take down their advice, it was also easier for them to inter-
the issues involved. There were a number of key issues within
                                                                 view others (as opposed to discussing their own ideas with
the survey questions that were repeated often, such as time
                                                                 interviewees).
frames, the use of singular or plural ‘you’ and terms such as
‘neighbours’, ‘friends’ and ‘family’. Individual interviews
were very helpful in understanding interpretations but did       Step 4: Reviewing questions for inclusion, revising the
not give us the opportunity to see opposing suggestions          question or intended meaning, or dropping questions
debated. We therefore held group discussions on focused
                                                                 The field transcripts and results from each research method
issues. The group discussions lasted 1–2 hours, and were led
                                                                 were collated and summarized by the authors for each survey
by two moderators. Each interview was recorded and later
                                                                 question. Reviews of questions were supported by a dis-
transcribed. Each group involved people with similar ages
                                                                 cussion between the authors and field assistants. Following
and gender, and the sum of groups captured the range of ages
                                                                 these discussions, decisions were made concerning the action
and gender in the area. Box 6 gives an example of our
                                                                 required for each question including:
approach to finding out how people interpreted the word
‘homestead’.
                                                                 • acceptance of the original question and meaning;
                                                                 • acceptance of the original question, with a change to the
                                                                   meaning;
Expert evaluation
                                                                 • changing the question (slightly) but keeping the meaning
Expert evaluation involved no interaction with potential           the same;
survey respondents, but rather those who could offer particu-    • dropping the question;
lar guidance on specific issues. The ‘experts’ included two       • writing a new question (and new intended meaning).
groups of people: external advisors including Kikamba Bible
translators and Kamba university lecturers on sociology; and     Any changes were tested through additional expanded
the field assistants living locally in Makueni who had assisted   interviews and any new questions also went through various
the KENQOL research group for several years. The external        translation checks. Prior to accepting any new or changed
language advisors reviewed the written survey questions and      questions all data on the question were once again collated
focused particularly on their interpretation of the meaning of   and reviewed prior to being taken forward to piloting.
328                                              How to do (or not to do) . . .


Piloting                                                           into other languages. There has been some use of qualitative
                                                                   techniques to assess the content validity of original (e.g.
Moser and Kalton (1992) refer to piloting as the ‘dress
                                                                   Saxena et al. 1998), but more usually the adapted versions of,
rehearsal’. Questions are placed together as it is expected that
                                                                   generic HRQL instruments (e.g. Chwalow et al. 1992; Lam et
they will appear on the final questionnaire and the dynamics of
                                                                   al. 1994), as well as the applicability of generic instruments to
the survey as a whole are investigated. The KENQOL group
                                                                   specific disease groups (e.g. Tazaki et al. 1998). These investi-
undertook piloting in two stages. The first focused on ques-
                                                                   gations tend to be relatively unstructured, and can be
tions from each sub-section (n = 6) of the survey. The second
                                                                   described as cursory investigations. In general there is
involved straight runs through the entire survey. One of the
                                                                   too much emphasis placed on quantitative assessments of
authors was present at each of these interviews, and a field
                                                                   reliability, validity and sensitivity. The quantitative assess-
assistant conducted the interview in Kikamba. At both stages
                                                                   ment is a rather closed approach, that ignores the respon-
of piloting, the interviewer noted down the response to each
                                                                   dent’s own interpretation of questions.
question (yes, no, don’t know, no response). They also noted
any other information the respondent gave, such as asking for
                                                                   We therefore adopted a qualitative approach to the assess-
clarification of the question, or the respondent providing
                                                                   ment of validity, reliability and sensitivity. This made both the
additional information after responding ‘yes’, ‘no’ or ‘don’t
                                                                   researcher’s and the respondent’s interpretation of questions
know’. In addition, the authors noted any body language,
                                                                   more open and accountable, and ensured that the final survey
pauses before a response, emotional reactions to questions etc.
                                                                   questions were more reliable and valid and able to detect
At the end of the pilot survey interviews, each respondent was
                                                                   change as revisions were made. Earlier we defined a reliable
asked for their opinion about the questions, including:
                                                                   measure as one which yields the same results in repeated
                                                                   trials (Carmine and Zeller 1979). A question may be un-
• what they thought about the questions in general;
                                                                   reliable due to bad wording, causing a person to understand
• whether any of the questions seemed to be strange or
                                                                   the question differently on different occasions (De Vaus
  unusual;
                                                                   1995). This paper has described procedures for assessing the
• their opinion on the order of questions;
                                                                   reliability of questions during the pre-testing stage, for
• how appropriate the response categories were;
                                                                   example identifying questions that cannot elicit a double
• whether the correct procedures are used to address:
                                                                   negative for a response, do not contain high level or old fash-
  (a) married and unmarried people,
                                                                   ioned language, and are free from jargon. A valid measure is
  (b) old and young people;
                                                                   one which measures what is intended (Carmine and Zeller
• whether any questions should not be asked in the survey;
                                                                   1979), and methods described in this paper which helped to
• to point out any questions they do not want to answer, or
                                                                   address this included ensuring that the questions capture
  think they should not be asked;
                                                                   current views, reflect local issues and the conception of health
• any problems they thought we might encounter asking
                                                                   each question represents is clear. Similarly, during pre-testing
  people the questions:
                                                                   the sensitivity to questions over time was investigated. De
  (a) alone,
                                                                   Vaus (1995) explains that instead of this procedure some
  (b) with others listening (e.g. young people were asked
                                                                   people include a large number of indicators and at a later
       whether they might have inhibitions answering ques-
                                                                   stage eliminate those that do not seem to be working.
       tions in front of parents);
                                                                   However, he explains: ‘this seems to be the wrong way of
• whether any questions appeared to be asking the same
                                                                   doing things since we will end up by defining the concept in
  thing;
                                                                   terms of the indicators which ‘worked’. If this is done then the
• their opinion on what changes we could make to improve
                                                                   indicators may not represent the concepts of the theory we
  this survey in terms of:
                                                                   set out to test and as such the research can end up having
  (a) introduction,
                                                                   little relevance to the original research question’ (De Vaus
  (b) questions;
                                                                   1995: 54).
• which questions might be difficult for someone to talk
  about in front of an interviewer who comes from the same
                                                                   However, we recognize that the process was labour intensive
  area as the respondent;
                                                                   and took a considerable amount of time (around 6 months of
• whether it matters who interviews people;
                                                                   researcher and 12 months of field assistant’s time). In our case
• who would be a good interviewer in terms of age, gender
                                                                   we judged this to be an important part of the process and
  and any other characteristics.
                                                                   outcome of our research. First, we are intending that this
                                                                   measure becomes used across many different diseases as
Based on feedback during piloting, several additional
                                                                   an outcome measure used to judge the effectiveness and
revisions were made to the KENQOL questions. Each
                                                                   efficiency of health-improving interventions. It is therefore
change was followed up by further piloting (and pre-testing
                                                                   important to us that the people whose health we are trying to
where necessary) until the final version of the KENQOL
                                                                   measure (and improve) are integrated into this process, and
survey was agreed.
                                                                   it is often the case that generic HRQL instruments are devel-
                                                                   oped and used for years following their development. Sec-
                                                                   ondly, it provided valuable comparative data for explaining
Discussion
                                                                   any similarities and dissonances in measured health in Vihiga
We have been concerned about the procedures used in the            district, Kenya. Related to this is the possibility that others
development of HRQL measures as well as their translations         would want to translate the instrument for use elsewhere, and
How to do (or not to do) . . .                                                      329

we wanted to improve the basis from which judgements are          required information through household survey, readers
made about the equivalence of future translations.                should consider some of the literature from demographic
                                                                  studies; some examples include Caldwell and Igun (1997),
There are other techniques that are available to researchers      Ewbank (1981), Bledsoe and Pison (1994). We found cogni-
to test their survey questions that we did not use. The           tive research assessment a particularly inspiring approach as
KENQOL researchers adopted some cognitive assessment              it captures approaches given to a variety of psychological
survey methodologies, but there are others that may be more       processes involved in information retrieval. Some of this
appropriate for other research groups (Jobe and Mingay            literature has already been cited above, but readers might
1989):                                                            also like to consider Lessler and Tourangeu (1989) and the
                                                                  proceedings from a seminar on the Cognitive Aspects of
• Concurrent think-aloud interviews: respondents think            Survey Methodology (Sirken et al. 1999). Researchers in the
  aloud when answering questions, and responses are               field of marketing have also been considering these issues for
  probed.                                                         some time, consider for example Gendall and Hoek (1990)
• Retrospective think-aloud interviews: respondents answer        and Kumar et al. (1999).
  all question first and then are asked how they arrived at
  their answers.
• Confidence ratings: respondents relate the degree of confi-       Conclusions/recommendations
  dence they have in the accuracy of their answers.               The investigations described in this paper should not be seen
• Paraphrasing: respondents repeat the questions in their         as an unusual step in the development and testing of survey
  own words.                                                      questionnaires – particularly those that will be used repeat-
• Response latency: measurement of the elapsed time               edly. The validity and reliability (and possibly sensitivity) of
  between the questions and respondent’s answers.                 any survey is important to establish prior to accepting results
                                                                  and conclusions. Each of these ideas can begin to be explored,
The KENQOL group also adopted some methods recognized             and improved, through the use of qualitative methods in pre-
as participatory or rapid rural appraisal (PRA and RRA)           testing and piloting questionnaires. Indeed, evidence that a
approaches, for example the use of anthropologists and local      survey tool has ‘content validity’ is required, rather than
leaders as key informants (Chambers 1983). Other useful           simply relying on unsubstantiated assertions by researchers,
RRA and PRA approaches might include a well-being (or             and only once this has been demonstrated is it appropriate to
HRQL) ranking exercise of local households with local             adopt quantitative assessment methods for further analysis.
participants, and linking this to likely responses to questions   Finally, the knowledge gained about the interpretation of
and analysis of difference based on responses to questions        survey questions will also help in judging the equivalence
(see Chambers 1997).                                              of any future translations of the survey questionnaire.

Whilst the approach we have taken to the qualitative assess-
ment of meaning attached to the KENQOL survey questions           Endnotes
is unique amongst generic HRQL research, many of the ideas              1 Validity determines that an instrument is measuring what was
were drawn from a variety of sources, including survey            intended. It concerns the crucial relationship between concept and
research (e.g. Moser and Kalton 1992; Fowler 1993), trans-        indicator. The objective is to ensure that the indicator represents the
lation studies (e.g. Larson 1984; Barnwell 1992), the cognitive   intended (and only the intended) concept (Carmines and Zeller
sciences (e.g. Jobe 1990; Forsyth and Lessler 1991) as well as    1979).
                                                                        2 Reliability concerns the extent to which an experiment, test
the health measurement literature (e.g. Guillemin et al. 1993;    and any measuring procedure yields the same results on repeated
Herdman et al. 1997, 1998). De Vaus (1995) is particularly        trials (Carmines and Zeller 1979).
useful because of the links made between investigating ques-            3 The ability of a measure to detect change.

tionnaire interpretation and the reliability and validity of a          4 Because most household interviews happen outside the home

questionnaire. This gives an indication of the applicability of   with many people around. To ask to interview people privately in a
our approach to the design of survey questions more gener-        household survey tends to arouse a lot of suspicion.
                                                                        5 Our household survey eventually consisted of 112 health
ally. As Tourangeau (1984) states ‘[in] the respondent’s task
                                                                  measurement questions and 50 socioeconomic and demographic
. . . there is considerable room for error. Respondents may       questions. We achieved a total of 550 interviews – a response rate of
misunderstand the question or the response categories; they       81.3% and a mean item response rate of 98.1% (SD 3.6).
may forget or misremember the crucial information; they may             6 All interviews were undertaken in Kikamba, the local lan-

misjudge the information they do recall; and they may mis-        guage of the Kamba population we were working amongst. The
report their answer’ (Tourangeau 1984: 73–74).                    Kikamba versions of the KENQOL questions were used for all
                                                                  survey testing, although the English versions are reported here. The
                                                                  pre-test interviews were written up in full in English to facilitate
Finally, to help researchers give greater attention to investi-   participation by the non-Kikamba speaking researchers.
gating whether respondents interpret questions in ways                  7 wailu, wianie, utheu, useo, wiwa nesa, uima.
researchers expect (or hope), we give a brief introduction to
a variety of different types of source texts. Most social
research texts address questionnaire design, but few provide      References
detail on investigations into question meanings. One text that    Barnwell K. 1992. Bible Translation: An introductory course in trans-
is useful in this context is Oppenheim (2000). For an excellent       lation principles (3rd edition). Dallas, TX: Summer Institute of
insight into the many pitfalls associated with eliciting the          Linguistics International.
330                                                    How to do (or not to do) . . .

Bledsoe C, Pison G. 1994. Introduction. In: Bledsoe C, Pison G (eds).     Lessler JT, Tourangeau R. 1989. Vital health statistics. Series 6.1.
     Nuptiality in Sub-Saharan Africa: Contemporary anthropo-                  Washington, DC: National Center for Health Statistics.
     logical and demographic perspectives. Oxford: Clarendon Press,       Moser CA, Kalton G. 1992. Survey methods in social investigation
     pp. 1–22.                                                                 (2nd edition). Aldershot, UK: Gower.
Bowden A, Fox-Rushby J. 2000. A critical review of the adaptation         Oppenheim AN. 2000. Questionnaire design, interviewing and
     and use of generic HRQL measures amongst populations                      attitude measurement. Buckingham, UK: Open University
     outside of North America, western/southern/northern Europe,               Press.
     Australia and New Zealand. Quality of Life Research 9: 313.          Saxena S, Chandiramani K, Bhargava R. 1998. WHOQOL-Hindi: A
Caldwell JC, Igun AA. 1971. An experiment with census-type age                 questionnaire for assessing the quality of life in health care set-
     enumeration in Nigeria. Population Studies 25: 287–302.                   tings in India. National Medical Journal of India 11: 160–5.
Carmines EG, Zeller RA. 1979. Reliability and validity assessment.        Sirken M, Jabine T, Willis G, Martin E, Tucker C. 1999. A New
     London: Sage Publications.                                                Agenda for Interdisciplinary Survey Research Methods: Pro-
Chambers R. 1983. Rural development: putting the last first. Harlow,            ceedings of the CASM II Seminar. Hyattsville, MD: National
     UK: Longman.                                                              Center for Health Statistics. [http://www.cdc.gov/nchs/data/
Chambers R. 1997. Whose reality counts? Putting the first last.                 casm2pro.pdf, 2nd August 2001]
     London: Intermediate Technology.                                     Tazaki M, Nakane Y, Endo T et al. 1998. Results of a qualitative and
Chwalow AJ, Lurie A, Bean K et al. 1992. A French version of the               field study using the WHOQOL instrument for cancer patients.
     Sickness Impact Profile (SIP): stages in the cross-cultural vali-          Japanese Journal of Clinical Oncology 28: 134–41.
     dation of a generic quality of life scale. Fundamental Clinical      Tourangeau R. 1984. Cognitive sciences and survey methods. In:
     Pharmacology 6: 319–26.                                                   Jabine TB (ed). Cognitive aspects of survey methodology: build-
De Vaus. 1995. Surveys in social research (4th edition). London:               ing a bridge between disciplines. Washington, DC: National
     Routledge.                                                                Academy Press, pp. 73–100.
Ewbank DC. 1981. Age misreporting and age-selective undernumer-
     ation: sources, patterns, and consequences for demographic
     analysis. Committee on Population and Demography. Report             Acknowledgements
     No. 4. Washington, DC: National Academy Press.
Forsyth BH, Lessler JT. 1991. Cognitive laboratory methods: a tax-        We thank the following organizations for the funding of this survey:
     onomy. In: Biemer PP, Groves SM, Lyberg LE, Mathiowetz               UNDP/World Bank/WHO Special Programme for Research and
     NA, Sudan S (eds). Measurement errors in surveys. New York:          Training in Tropical Disease, the UK Economic and Social Research
     John Wiley, pp. 393–418.                                             Council (Ref. H52427005994) and the Health Economics and
Fowler FJ. 1993. Survey research methods (2nd edition). London:           Financing Programme (funded by the Department for International
     Sage Publications.                                                   Development). We also thank all our field assistants, especially
Fox-Rushby J, Parker M. 1995. Culture and the measurement of              Justus Mutunga, Jeremiah Nyamai and Alex Muvea, for their help,
     health related quality of life. European Review of Applied Psy-      guidance and enthusiasm. Finally we thank all the people who
     chology 45: 257–63.                                                  participated in the preparation for this survey and made us welcome
Fox-Rushby J, Johnson K, Mwanzo I et al. 1997. Creating an instru-        in their homes.
     ment to assess lay perceptions of HRQL: options and impli-
     cations. Quality of Life Research 6: 633.
Fox-Rushby J, for the KENQOL Group. 2000. Operationalising con-           Biographies
     ceptions of ‘health’ amongst the Wakamba and Maragoli of             Annabel Bowden, Ph.D. (Health Geography), was a Research Fellow
     Kenya: the basis of the KENQOL instrument. Quality of Life           at the London School of Hygiene and Tropical Medicine (LSHTM)
     Research 9: 316.                                                     at the time of this study. She was involved with the KENQOL project
Gendall P, Hoek J. 1990. A question of wording. Marketing Bulletin        for 5 years and was based in Kenya for 1 year during the household
     1: 25–36.                                                            survey. She also has a wider interest in the applicability of
Grosh M, Glewwe P. 2000. Designing household survey question-             adapted/translated versions of generic health-related quality of life
     naires for developing countries: lessons from 15 years of the        measures. She is now based with the Lewin Group (UK).
     Living Standard Measurement Survey (Volume 1). Washington,
     DC: World Bank.
                                                                          Julia Fox-Rushby, Ph.D. (Economics), is a Senior Lecturer at the
Guillemin F, Bombardier C, Beaton D. 1993. Cross-cultural adap-
                                                                          LSHTM. She has written numerous academic papers on the cost-
     tation of health-related quality of life measures: literature
                                                                          effectiveness of health interventions across the world, specializing
     review and proposed guidelines. Journal of Clinical Epidemiol-
                                                                          particularly in maternal and child health, malaria and, more recently,
     ogy 46: 1417–32.
                                                                          vaccine preventable disease. She has also been involved over the past
Herdman M, Fox-Rushby JA, Badia X. 1997. ‘Equivalence’ and the
                                                                          15 years in developing a number of non-disease specific measures of
     translation and adaptation of health-related quality of life ques-
                                                                          health-related quality of life as a member of the EuroQol group,
     tionnaires. Quality of Life Research 6: 237–47.
                                                                          advisor to the WHOQOL group and as principal investigator of the
Herdman M, Fox-Rushby JA, Badia X. 1998. A model of equival-
                                                                          KENQOL group.
     ence in the cultural adaptation of HRQL measures: a univer-
     salist approach. Quality of Life Research 7: 323–35.
Jobe JB. 1990. Research on questionnaire design: perspectives from        Lilian Nyandieka, MA (Sociology), is a Research Officer at the
     other disciplines. American Journal of Epidemiology 132: 824.        Center for Public Health Research, a department of the Kenya
Jobe JB, Mingay DJ. 1989. Cognitive research improves question-           Medical Research Institute (PO Box 20752, Nairobi, Kenya). She has
     naires. American Journal Public Health 79: 1053–5.                   been involved in the KENQOL project for 3 years and is also
Kenya Government. 2000. Makueni District Development Report               researching social and economic aspects of malaria.
     1997–2001. Nairobi: Office of the Vice President and Ministry of
     Planning and National Development.                                   John Wanjau, BA (Anthropology), is an Assistant Research Officer
Kumar V, Aaker DA, Day GS. 1999. Essentials of marketing                  at the Center for Public Health Research, Kenya Medical Research
     research. Chichester, UK: John Wiley.                                Institute. He has been involved in the KENQOL project for 3 years.
Lam CL, Van Weel C, Lauder IJ. 1994. Can the Dartmouth COOP/
     WONCA charts be used to assess the functional status of              Correspondence: Julia Fox-Rushby Ph.D., Health Policy Unit,
     Chinese patients? Family Practice 11: 85–94.                         London School of Hygiene and Tropical Medicine, Keppel Street,
Larson ML. 1984. Meaning based translation: a guide to cross-lan-         London WC1E 7HT, UK. Tel: +44 (0) 20 7927 2267, Fax: +44 (0) 20
     guage equivalence. Lanham, MD: University Press of America           7637 5391, Email: Julia.Fox-Rushby@lshtm.ac.uk

More Related Content

What's hot

L. O'Keefe Writing Sample - Qualitative Research Narrative
L. O'Keefe Writing Sample - Qualitative Research NarrativeL. O'Keefe Writing Sample - Qualitative Research Narrative
L. O'Keefe Writing Sample - Qualitative Research NarrativeLindsay O'Keefe
 
Research Beyond the Theory
Research Beyond the TheoryResearch Beyond the Theory
Research Beyond the TheoryHazemGhaith
 
Clinical practice critical_research_paper essay sample from assignmentsupport...
Clinical practice critical_research_paper essay sample from assignmentsupport...Clinical practice critical_research_paper essay sample from assignmentsupport...
Clinical practice critical_research_paper essay sample from assignmentsupport...https://writeessayuk.com/
 
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVE
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVEMETHODS OF RESEARCH QUALITATIVE AND QUANTITATIVE
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVEswarna2912
 
Research in Nursing: A Guide to Understanding Research Designs and Techniques
Research in Nursing: A Guide to Understanding Research Designs and TechniquesResearch in Nursing: A Guide to Understanding Research Designs and Techniques
Research in Nursing: A Guide to Understanding Research Designs and TechniquesAJHSSR Journal
 
Bayesian Model for Multivatiate Functional Principle Components Analysis
Bayesian Model for Multivatiate Functional Principle Components AnalysisBayesian Model for Multivatiate Functional Principle Components Analysis
Bayesian Model for Multivatiate Functional Principle Components AnalysisKevin Cummins
 
factor influencing relapse mental disorders.pdf
factor influencing relapse mental disorders.pdffactor influencing relapse mental disorders.pdf
factor influencing relapse mental disorders.pdfYuaKim
 
GEMC - Nursing Research
GEMC - Nursing ResearchGEMC - Nursing Research
GEMC - Nursing ResearchOpen.Michigan
 
Qualitative methods in Health techcnology assessment
Qualitative methods in Health techcnology assessmentQualitative methods in Health techcnology assessment
Qualitative methods in Health techcnology assessmentLevis Kahandukya Nyavanda
 
Ebp lec ppt june 2012
Ebp lec ppt june 2012Ebp lec ppt june 2012
Ebp lec ppt june 2012mappsanchez
 
Importance across fields
Importance across fieldsImportance across fields
Importance across fieldsmrsportes18
 
Research methodology module-2
Research methodology module-2Research methodology module-2
Research methodology module-2Satyajit Behera
 
Research Methods in Psychology
Research Methods in PsychologyResearch Methods in Psychology
Research Methods in PsychologyRegent University
 
Research methodology module-1
Research methodology module-1Research methodology module-1
Research methodology module-1Satyajit Behera
 

What's hot (17)

L. O'Keefe Writing Sample - Qualitative Research Narrative
L. O'Keefe Writing Sample - Qualitative Research NarrativeL. O'Keefe Writing Sample - Qualitative Research Narrative
L. O'Keefe Writing Sample - Qualitative Research Narrative
 
Research Beyond the Theory
Research Beyond the TheoryResearch Beyond the Theory
Research Beyond the Theory
 
Clinical practice critical_research_paper essay sample from assignmentsupport...
Clinical practice critical_research_paper essay sample from assignmentsupport...Clinical practice critical_research_paper essay sample from assignmentsupport...
Clinical practice critical_research_paper essay sample from assignmentsupport...
 
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVE
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVEMETHODS OF RESEARCH QUALITATIVE AND QUANTITATIVE
METHODS OF RESEARCH QUALITATIVE AND QUANTITATIVE
 
Research in Nursing: A Guide to Understanding Research Designs and Techniques
Research in Nursing: A Guide to Understanding Research Designs and TechniquesResearch in Nursing: A Guide to Understanding Research Designs and Techniques
Research in Nursing: A Guide to Understanding Research Designs and Techniques
 
Bayesian Model for Multivatiate Functional Principle Components Analysis
Bayesian Model for Multivatiate Functional Principle Components AnalysisBayesian Model for Multivatiate Functional Principle Components Analysis
Bayesian Model for Multivatiate Functional Principle Components Analysis
 
factor influencing relapse mental disorders.pdf
factor influencing relapse mental disorders.pdffactor influencing relapse mental disorders.pdf
factor influencing relapse mental disorders.pdf
 
Proposal writing by dr.s.kalpana
Proposal writing by dr.s.kalpanaProposal writing by dr.s.kalpana
Proposal writing by dr.s.kalpana
 
GEMC - Nursing Research
GEMC - Nursing ResearchGEMC - Nursing Research
GEMC - Nursing Research
 
Qualitative methods in Health techcnology assessment
Qualitative methods in Health techcnology assessmentQualitative methods in Health techcnology assessment
Qualitative methods in Health techcnology assessment
 
Ebp lec ppt june 2012
Ebp lec ppt june 2012Ebp lec ppt june 2012
Ebp lec ppt june 2012
 
Importance across fields
Importance across fieldsImportance across fields
Importance across fields
 
Qualitative and descriptive research
Qualitative and descriptive researchQualitative and descriptive research
Qualitative and descriptive research
 
Research methodology module-2
Research methodology module-2Research methodology module-2
Research methodology module-2
 
Class 2, Final
Class 2, FinalClass 2, Final
Class 2, Final
 
Research Methods in Psychology
Research Methods in PsychologyResearch Methods in Psychology
Research Methods in Psychology
 
Research methodology module-1
Research methodology module-1Research methodology module-1
Research methodology module-1
 

Viewers also liked

An alternative method for meta analysis
An alternative method for meta analysisAn alternative method for meta analysis
An alternative method for meta analysisrsd kol abundjani
 
Methodological issues in grounded theory
Methodological issues in grounded theoryMethodological issues in grounded theory
Methodological issues in grounded theoryrsd kol abundjani
 
A simple confidence interval for meta analysis
A simple confidence interval for meta analysisA simple confidence interval for meta analysis
A simple confidence interval for meta analysisrsd kol abundjani
 
A multiple case study of implementation ijn 10 loca
A multiple case study of implementation ijn 10 locaA multiple case study of implementation ijn 10 loca
A multiple case study of implementation ijn 10 locarsd kol abundjani
 
Combining the analysies of three qualitative data
Combining the analysies of three qualitative data Combining the analysies of three qualitative data
Combining the analysies of three qualitative data rsd kol abundjani
 
The approriate uses of qualitative methods
The approriate uses of qualitative methodsThe approriate uses of qualitative methods
The approriate uses of qualitative methodsrsd kol abundjani
 
Hubungan peng kes dengan perilaku hiegenes menstruasi
Hubungan peng kes dengan perilaku hiegenes menstruasiHubungan peng kes dengan perilaku hiegenes menstruasi
Hubungan peng kes dengan perilaku hiegenes menstruasirsd kol abundjani
 
Penelitian deskriptif analitis (sulipan)
Penelitian deskriptif analitis (sulipan)Penelitian deskriptif analitis (sulipan)
Penelitian deskriptif analitis (sulipan)rsd kol abundjani
 
Using simulation in out patient queues a case study
Using simulation in out patient queues a case studyUsing simulation in out patient queues a case study
Using simulation in out patient queues a case studyrsd kol abundjani
 
Asssssssssssssssssssssssssssssssssaaaaaaaaa
AsssssssssssssssssssssssssssssssssaaaaaaaaaAsssssssssssssssssssssssssssssssssaaaaaaaaa
Asssssssssssssssssssssssssssssssssaaaaaaaaarsd kol abundjani
 
Leadership style and perceived benefits of electronic data interchange for re...
Leadership style and perceived benefits of electronic data interchange for re...Leadership style and perceived benefits of electronic data interchange for re...
Leadership style and perceived benefits of electronic data interchange for re...rsd kol abundjani
 

Viewers also liked (19)

An alternative method for meta analysis
An alternative method for meta analysisAn alternative method for meta analysis
An alternative method for meta analysis
 
Methodological issues in grounded theory
Methodological issues in grounded theoryMethodological issues in grounded theory
Methodological issues in grounded theory
 
Naskah+publikasi+rita
Naskah+publikasi+ritaNaskah+publikasi+rita
Naskah+publikasi+rita
 
A simple confidence interval for meta analysis
A simple confidence interval for meta analysisA simple confidence interval for meta analysis
A simple confidence interval for meta analysis
 
langkah metode ilmiah
langkah metode ilmiahlangkah metode ilmiah
langkah metode ilmiah
 
A multiple case study of implementation ijn 10 loca
A multiple case study of implementation ijn 10 locaA multiple case study of implementation ijn 10 loca
A multiple case study of implementation ijn 10 loca
 
Combining the analysies of three qualitative data
Combining the analysies of three qualitative data Combining the analysies of three qualitative data
Combining the analysies of three qualitative data
 
Besar sampel b
Besar sampel bBesar sampel b
Besar sampel b
 
The approriate uses of qualitative methods
The approriate uses of qualitative methodsThe approriate uses of qualitative methods
The approriate uses of qualitative methods
 
Berhubungan marbu
Berhubungan marbuBerhubungan marbu
Berhubungan marbu
 
Hubungan peng kes dengan perilaku hiegenes menstruasi
Hubungan peng kes dengan perilaku hiegenes menstruasiHubungan peng kes dengan perilaku hiegenes menstruasi
Hubungan peng kes dengan perilaku hiegenes menstruasi
 
Penelitian deskriptif analitis (sulipan)
Penelitian deskriptif analitis (sulipan)Penelitian deskriptif analitis (sulipan)
Penelitian deskriptif analitis (sulipan)
 
Using simulation in out patient queues a case study
Using simulation in out patient queues a case studyUsing simulation in out patient queues a case study
Using simulation in out patient queues a case study
 
7
77
7
 
Asssssssssssssssssssssssssssssssssaaaaaaaaa
AsssssssssssssssssssssssssssssssssaaaaaaaaaAsssssssssssssssssssssssssssssssssaaaaaaaaa
Asssssssssssssssssssssssssssssssssaaaaaaaaa
 
Leadership style and perceived benefits of electronic data interchange for re...
Leadership style and perceived benefits of electronic data interchange for re...Leadership style and perceived benefits of electronic data interchange for re...
Leadership style and perceived benefits of electronic data interchange for re...
 
Kadarzi
KadarziKadarzi
Kadarzi
 
Standar persyaratan minimal
Standar persyaratan minimalStandar persyaratan minimal
Standar persyaratan minimal
 
Korelasional spss1
Korelasional spss1Korelasional spss1
Korelasional spss1
 

Similar to [282 met] 322

Version 2 Critique of a Quantitative Research.docx
Version 2 Critique of a Quantitative Research.docxVersion 2 Critique of a Quantitative Research.docx
Version 2 Critique of a Quantitative Research.docxwrite22
 
Qualitative Research Methods for Public Health
Qualitative Research Methods for Public HealthQualitative Research Methods for Public Health
Qualitative Research Methods for Public HealthCelia Emmelhainz
 
Foundations of research methodology for Post Graduate students by Dr.KKK.pptx
Foundations of research methodology for Post Graduate students by Dr.KKK.pptxFoundations of research methodology for Post Graduate students by Dr.KKK.pptx
Foundations of research methodology for Post Graduate students by Dr.KKK.pptxKatareKiranKumar
 
Chapter2-Methods_of_Research-Module.pdf
Chapter2-Methods_of_Research-Module.pdfChapter2-Methods_of_Research-Module.pdf
Chapter2-Methods_of_Research-Module.pdfJUNGERONA
 
Stepby-step guide to critiquingresearch. Part 1 quantitati.docx
Stepby-step guide to critiquingresearch. Part 1 quantitati.docxStepby-step guide to critiquingresearch. Part 1 quantitati.docx
Stepby-step guide to critiquingresearch. Part 1 quantitati.docxsusanschei
 
The Purpose Statement, Research Questions, and Hypothesis
The Purpose Statement, Research Questions, and HypothesisThe Purpose Statement, Research Questions, and Hypothesis
The Purpose Statement, Research Questions, and Hypothesisnncygarcia
 
Introduction to Research.pptx
Introduction to Research.pptxIntroduction to Research.pptx
Introduction to Research.pptxCheriro
 
Researchmethodology 110218234528-phpapp02
Researchmethodology 110218234528-phpapp02Researchmethodology 110218234528-phpapp02
Researchmethodology 110218234528-phpapp02Hamizo
 
Research methodology
Research methodologyResearch methodology
Research methodologyAshish Sahu
 
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docx
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docxAuthor & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docx
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docxrock73
 
Group twob ppt_presentation_1
Group twob ppt_presentation_1Group twob ppt_presentation_1
Group twob ppt_presentation_1grouptwob
 
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...MEASURE Evaluation
 
Research methodology
Research methodologyResearch methodology
Research methodologyMATEEN YOUNIS
 

Similar to [282 met] 322 (20)

Version 2 Critique of a Quantitative Research.docx
Version 2 Critique of a Quantitative Research.docxVersion 2 Critique of a Quantitative Research.docx
Version 2 Critique of a Quantitative Research.docx
 
Qualitative Research Methods for Public Health
Qualitative Research Methods for Public HealthQualitative Research Methods for Public Health
Qualitative Research Methods for Public Health
 
Foundations of research methodology for Post Graduate students by Dr.KKK.pptx
Foundations of research methodology for Post Graduate students by Dr.KKK.pptxFoundations of research methodology for Post Graduate students by Dr.KKK.pptx
Foundations of research methodology for Post Graduate students by Dr.KKK.pptx
 
qualmethods.pdf
qualmethods.pdfqualmethods.pdf
qualmethods.pdf
 
Chapter2-Methods_of_Research-Module.pdf
Chapter2-Methods_of_Research-Module.pdfChapter2-Methods_of_Research-Module.pdf
Chapter2-Methods_of_Research-Module.pdf
 
Stepby-step guide to critiquingresearch. Part 1 quantitati.docx
Stepby-step guide to critiquingresearch. Part 1 quantitati.docxStepby-step guide to critiquingresearch. Part 1 quantitati.docx
Stepby-step guide to critiquingresearch. Part 1 quantitati.docx
 
The Purpose Statement, Research Questions, and Hypothesis
The Purpose Statement, Research Questions, and HypothesisThe Purpose Statement, Research Questions, and Hypothesis
The Purpose Statement, Research Questions, and Hypothesis
 
[276 met]
[276 met][276 met]
[276 met]
 
Introduction to Research.pptx
Introduction to Research.pptxIntroduction to Research.pptx
Introduction to Research.pptx
 
Researchmethodology 110218234528-phpapp02
Researchmethodology 110218234528-phpapp02Researchmethodology 110218234528-phpapp02
Researchmethodology 110218234528-phpapp02
 
Research methodology
Research methodologyResearch methodology
Research methodology
 
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docx
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docxAuthor & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docx
Author & TitleAuthors Maggie Lawrence & Sue Kinn.Title Need.docx
 
Group 1.pdf
Group 1.pdfGroup 1.pdf
Group 1.pdf
 
Group twob ppt_presentation_1
Group twob ppt_presentation_1Group twob ppt_presentation_1
Group twob ppt_presentation_1
 
Qualitative research methods for student
Qualitative research methods for studentQualitative research methods for student
Qualitative research methods for student
 
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...
Using Cognitive Interviews in Nigeria to Pre-test Child, Caregiver, and House...
 
Qualitative Essay
Qualitative EssayQualitative Essay
Qualitative Essay
 
Research methodology
Research methodologyResearch methodology
Research methodology
 
Research methodology
Research methodologyResearch methodology
Research methodology
 
Research methodology
Research methodologyResearch methodology
Research methodology
 

More from rsd kol abundjani

More from rsd kol abundjani (20)

Rpkps
RpkpsRpkps
Rpkps
 
Modul 7-format-kpt
Modul 7-format-kptModul 7-format-kpt
Modul 7-format-kpt
 
Draft kurikulum-2013-per-tgl-13-november-2012-pukul-14
Draft kurikulum-2013-per-tgl-13-november-2012-pukul-14Draft kurikulum-2013-per-tgl-13-november-2012-pukul-14
Draft kurikulum-2013-per-tgl-13-november-2012-pukul-14
 
Aspek penilaian
Aspek penilaianAspek penilaian
Aspek penilaian
 
8. pengembangan bahan ajar
8. pengembangan bahan ajar8. pengembangan bahan ajar
8. pengembangan bahan ajar
 
Tema tema kkn-ppm1
Tema tema kkn-ppm1Tema tema kkn-ppm1
Tema tema kkn-ppm1
 
Tayang peranan wi dan tantangannya ddn 09-12-09
Tayang peranan wi dan tantangannya ddn 09-12-09Tayang peranan wi dan tantangannya ddn 09-12-09
Tayang peranan wi dan tantangannya ddn 09-12-09
 
Spmpt
SpmptSpmpt
Spmpt
 
Skd
SkdSkd
Skd
 
pengawasan mutu pangan
pengawasan mutu panganpengawasan mutu pangan
pengawasan mutu pangan
 
Rpp opd seminar executive edit
Rpp opd seminar executive editRpp opd seminar executive edit
Rpp opd seminar executive edit
 
Pelatihan applied approach
Pelatihan applied approachPelatihan applied approach
Pelatihan applied approach
 
Matematika bangun-datar
Matematika bangun-datarMatematika bangun-datar
Matematika bangun-datar
 
Kuliah pendahuluan bioo teknologi pertanian
Kuliah pendahuluan bioo teknologi pertanianKuliah pendahuluan bioo teknologi pertanian
Kuliah pendahuluan bioo teknologi pertanian
 
Konsep penulisan modul mata pelajaran
Konsep penulisan modul mata pelajaranKonsep penulisan modul mata pelajaran
Konsep penulisan modul mata pelajaran
 
Kerangka acuan dan laporan
Kerangka acuan dan laporanKerangka acuan dan laporan
Kerangka acuan dan laporan
 
Keindahan matematik dan angka
Keindahan matematik dan angkaKeindahan matematik dan angka
Keindahan matematik dan angka
 
Kebijakan nasional spmi pt
Kebijakan nasional spmi ptKebijakan nasional spmi pt
Kebijakan nasional spmi pt
 
Jurnal pelatihan jafung adminkes
Jurnal pelatihan jafung adminkesJurnal pelatihan jafung adminkes
Jurnal pelatihan jafung adminkes
 
Inventarisasi koleksi perpustakaan
Inventarisasi koleksi perpustakaanInventarisasi koleksi perpustakaan
Inventarisasi koleksi perpustakaan
 

[282 met] 322

  • 1. HEALTH POLICY AND PLANNING; 17(3): 322–330 © Oxford University Press 2002 How to do (or not to do) . . . Methods for pre-testing and piloting survey questions: illustrations from the KENQOL survey of health-related quality of life A BOWDEN,1 JA FOX-RUSHBY,1 L NYANDIEKA2 AND J WANJAU2 1Health Policy Unit, London School of Hygiene and Tropical Medicine, UK, and 2Centre for Public Health Research, Kenya Medical Research Institute, Nairobi, Kenya Part of the assertion that any survey researcher can make about the validity of their results needs to contain an analysis of questions and their responses from the respondent’s viewpoint. Claims concerning the valid- ity, reliability and sensitivity of health-related quality of life measures tend to be based on the quantitative approach of psychometrics, which fails to identify when respondents: misinterpret questions; do not recall the information requested; or give answers that present themselves in a better or worse light. The paper presents some approaches to pre-testing and piloting survey questionnaires to check the interpretation of survey questions, using illustrations from the KENQOL project. The paper describes: how the intended referential and connotative meaning of each question was established; the criteria to judge the appropriate- ness of each question; the methods used to make those judgements; and the process of reviewing questions based on findings. The role of piloting is highlighted, and further reading is suggested for readers wishing to develop a model for their own investigation. Key words: survey, pre-testing, piloting, Kenya, health-related quality of life Introduction We were in the process of developing a survey tool to measure the self-perceived and locally defined ‘health’ of Questionnaire-based surveys are a very common approach to people in Makueni District (Kenya), and reviewed practices data collection used by researchers and the interpretation of in the development and translation of such measures as part findings may be used to inform policy or public debate. It is of this process. We noticed that claims concerning the valid- not uncommon to find reports in the media saying ‘a recent ity, reliability and sensitivity of health-related quality of life survey showed that X% of people think . . . .’ But do they? (HRQL) measures tended to be based on the quantitative How much trust can we place in the findings of surveys? approach of psychometrics. Such quantitative techniques are Answering questions on a survey requires several thought unlikely to identify when respondents misinterpret ques- processes: respondents have to interpret the question, tions, fail to recall the information requested or give answers retrieve the necessary information and then decide what that present themselves in a better light. Indeed the interpre- answer to give – and all this before the researcher gives their tation of questionnaire items has often been overlooked own interpretation to the responses. (or at least not reported) in the development and trans- lation of HRQL measures (Fox-Rushby and Parker 1995; Part of the claim that any survey researcher can make about Bowden and Fox-Rushby 2000). It is also the case that well- the validity of their results needs to contain an analysis of the known texts in survey research often fail to address issues questions from the respondent’s viewpoint. This is a salient of how to pre-test survey questions in sufficient detail starting point either for those developing their own question- (Moser and Kalton 1992; Fowler 1993; Grosh and Glewwe naire or for those using an existing questionnaire in a new 2000). context (either in a new population or new language). Failing to investigate the interpretation of questionnaire items may The aim of this paper is to present a variety of examples to result in misinterpretations (by respondents and researchers), help researchers improve the validity,1 reliability2 and sensi- falsified answers, missing responses (to the questionnaire as a tivity3 of their survey instruments prior to undertaking a whole or to particular items), and possibly an offended household survey. The paper first provides a brief context to respondent who chases away the interviewer and encourages our research before proceeding to the range of methods used others to refuse interviews. to pre-test and pilot the survey instrument. The discussion
  • 2. How to do (or not to do) . . . 323 section focuses on the generalizability of our approach to Pre-testing the survey questions testing questions for surveys and introduces the reader to a Pre-testing is the main chance for researchers to gauge the range of literature for further reading. meaning attributed to survey questions ‘before it is too late’, i.e. before a substantial investment is made in the wrong Context of research questions or in questions where the researcher cannot be sure about what is being asked. It is therefore imperative that The overall aim of our research is to develop a culturally the researchers undertake these interviews, rather than relevant generic measure of ‘health’ to measure the impact of assigning the task to field assistants, because they know the interventions designed to reduce disease and/or improve aim of the research. The pre-testing involved a number of health in Kenya. We have been working amongst the Kamba steps including: of Makueni District in Eastern Kenya and the Maragoli of Vihiga District in Western Kenya. The conceptualization and • establishing the intended referential and connotative operationalization of ‘health’ emerged from extensive meaning of each question; periods of participant observation by anthropologists, in • agreeing a set of criteria to judge the appropriateness of addition to qualitative research by others (see Fox-Rushby et survey questions; al. 1997; Fox-Rushby, for the KENQOL Group, 2000). The • selecting the methods for judging appropriateness and definition of health adopted focused on the subjective per- undertaking research; ceptions of the positive and negative aspects of health that • reviewing questions for inclusion, revising (the question or comprised ‘contentment’, ‘cleanliness’, corporeal capacity’, intended meaning) or exclusion. ‘co-operation’ and ‘completeness’ of an intending individual within the intentional worlds of Makueni and Vihiga (Fox- Rushby, for the KENQOL Group, 2000). Step 1: Establishing the intended referential and connotative meaning of each question The KENQOL survey questionnaire was developed to It is important to ascertain whether respondents are inter- access those aspects of health that a member of the com- preting the question as intended to ensure appropriate munity could respond to in public4 to an interviewer who was interpretations are placed on the data. Therefore we pre- not necessarily known to them personally. This paper focuses pared descriptions of the intended referential and connota- on our work amongst the Kamba and reflects the develop- tive meaning for each of the survey questions drawn up. ment of the household survey in the Kamba language.5 The There were several advantages to this: it served as an account- aim of this part of our research was to investigate: how able approach to assessing content validity; it facilitated respondents understood the KENQOL survey questions; clarification of our own intentions in asking questions; it the types of information retrieved and over what time ensured we assessed the same hypotheses as there was less period; the types of socially acceptable answers that might variation in our own interpretation of the meaning of the be given and why; in addition to evaluating whether they original questions (either over time or between researchers); performed the task the way the researcher expected. We and it was expected to be helpful in the future assessment of tested a total of 150 potential health measurement survey the conceptual equivalence of the Kamba language version questions. with other language versions (see Herdman et al. 1997, 1998). Box 1 provides an example of the type of description we The field site was located in Makueni District in Eastern attached to a question. Kenya. It is approximately halfway between Nairobi and Mombasa, just north of the main trunk road. The Kamba population (part of the Central Bantu group) is the majority ethnic group in this region (99%) and traditionally they are pastoralists. Currently 50% of the population are aged 0–14 years but infant mortality rates are high, at 96 per 1000 live births in 1995 (Kenya Government 2000). Rainfall is Box 1. Sample meaning for a KENQOL survey question scarce in the area, with an annual precipitation of less than 500 mm and this, combined with high daytime temperatures, leads to a hot, dry environment. Agriculture is unreliable and Do you feel that there are some people who could sabo- food deficits are often experienced. Health facilities are rela- tage the progress of people in your neighbourhood at the tively scarce, although they can be reached by matatu from moment? the main Nairobi/Mombasa highway; the walk to the road can take up to 2 hours. We want to find out an individual’s current views about whether they feel that their ability to improve them- A day’s bus journey is required to reach hospital facilities in selves and their lives (collectively as a community) Nairobi. The major diseases are malaria, respiratory tract may be adversely affected by people within the com- infections, diarrhoeal diseases, skin ulcers, intestinal worms munity in some way. It is possible this may include a and urinary tract infections (Kenya Government 2000). range of ideas such as violence and witchcraft and we Incomes are low compared with many other areas in Kenya, leave it up to the respondent to decide if they are being with many people living below the poverty line and involved threatened. in a daily struggle for survival and development.
  • 3. 324 How to do (or not to do) . . . Table 1. Criteria selected for judging appropriateness of survey questions and investigative method employed 1 Criteria 2 Method Expanded Targeted Group Expert interviews methods methods reviews 1 No negative questions, and therefore no double negative answers √ √ 2 There are not two questions in one √ √ 3 Level of language is not too high/old fashioned/unusual in any sense √ √ 4 The question is simple and grammatically correct √ 5 The question is free from jargon √ 6 Singular and plural ‘you’ is clear √ √ √ 7 The same idea is not contained in another question √ √ 8 The question is sensitive to measure change over time √ √ 9 Captures current views √ √ 10 Reflects local issues √ √ 11 Meaning and interpretation of question is clear √ √ √ 12 The question makes sense to everyone √ √ 13 A ‘yes’ is clear and unambiguous in meaning √ √ 14 A ‘no’ is clear and unambiguous in meaning √ √ 15 Time period is clear √ √ √ 16 Conception(s) of health it represents is specified √ √ 17 How the question relates to: time of year; head of household; individual/ homestead/wider family/community √ √ √ 18 Understanding of how it relates to severity of HRQL √ √ Step 2: Criteria to judge the appropriateness of survey organize the expanded interviews we divided the set of survey questions questions into groups of between three and five questions. In each interview6 the respondent was first asked the KENQOL Having developed the survey question and agreed on the survey question and asked to respond either ‘yes’, ‘no’ or intended meaning, we developed a series of criteria for evalu- ‘don’t know’ as they chose. The probes used were divided into ating the appropriateness of the survey questions. This facili- three types and used in separate interviews: tated an agreed and standardized way of making decisions about which questions to include and enabled different (1) Comprehension of ‘key phrases’ within the question. We researchers to bring their own judgements together in a asked respondents to talk about the kinds of things they comparable way. Column 1 of Table 1 outlines the range of thought about when answering the question. Examples of criteria we adopted. It can be seen that the issues focus largely key phrases included the time frame the respondent on the style of language, attributed meaning, clarity of the considered and who respondents included when the question and consistency of interpretations. word ‘friends’ or ‘neighbours’ was used (see Box 2 for an example). (2) Applicability of the questions: to people of different ages, Step 3: Methods for judging appropriateness of survey gender and socioeconomic groups (see Box 3 for an questions There are a wide variety of techniques available to researchers wanting to test survey questions. We adopted four principal types (described in Forsyth and Lessler 1991): Box 2. Comprehension of key phrases in the question expanded interview techniques, targeted methods, group methods and expert evaluation. Others approaches are addressed in the discussion. Table 1 (column 2) indicates Do you feel safe at the moment? which technique(s) addresses which type of issue and this section outlines each method. Of these techniques, the The question intended to access concerns about per- expanded interviews took most of our time. sonal safety and safety of belongings. The probe ques- tions asked respondents to give examples for why someone (a) will feel safe and (b) may not feel safe. It Expanded interviews emerged that the question not only addressed notions Expanded interviews begin by one or more of the intended of personal safety and safety of belongings, but also survey questions being asked and answered by the respon- having enough food, being properly taken care of and dent, followed by a set of probe questions to investigate the having money for school fees. thought processes involved in answering the question(s). To
  • 4. How to do (or not to do) . . . 325 Box 3. Applicability of questions for respondents of different ages We asked respondents who in the family can and cannot be asked the question, ‘At the moment are you always worry- ing about getting food?’ The text below demonstrates some of our findings with respect to this question. Interview with an elderly man: Interviewer: Can children worry about getting food? Respondent: No, since it is their father’s fault they are fed. Interviewer: Could a young unmarried man answer this question? Respondent: A young unmarried man isn’t responsible for that. You will find that in most cases he comes at night and gets in his room and sleeps. Maybe he has passed to his friends’ home and had something to eat. The old mamas can’t worry since they are just looked after by their children they brought up. Interviewer: Is there anyone who might say no to this question? Respondent: Old and young people. example); by level of sensitivity of the question; and you ‘clean’? – meaning in a hygienic and moral sense), social desirability of responses (see Box 4). but further probing highlights some potential problems. Box 4 has a particularly interesting response because it (3) The meaning of a ‘yes’ or ‘no’ response needed to be clear also shows how sensitive the questioning on sensitivity and as intended (see Box 5). needs to be. It might at first appear that there are no prob- lems with the question ‘Ivindani yii nutonya kwiyikalya The respondents for the expanded interviews were volun- wi mutheu undu utonya kwenda?’ (At the moment, are teers willing to expend more time than the data collection Box 4. Addressing the sensitivity of the question We asked respondents directly about the sensitivity of questions. We needed to know whether a question was so sensi- tive that people might purposefully provide the wrong answer. Alternatively they may provide the right answer, but this makes the respondent or their family uncomfortable in some way. The transcript below is a translation from an inter- view that addressed sensitivity. Interviewer: At the moment, are you ‘clean’ [hygienically and morally]? Respondent: Yes Interviewer: When we ask this question in the interview, we want to be sure that we are not embarrassing the person we are asking. Do you think this question might cause embarrassment if it is asked in an interview? Respondent: The question is good and cannot cause embarrassment in an interview. Interviewer: Do you think people will feel they can answer this question truthfully in an interview? Respondent: Not all people can answer truthfully. This is because if one feels unclean or has done wrong, he wouldn’t want to be known to have wronged. Interviewer: If someone answers ‘no’ to this question, how will it reflect on the family of that person? Respondent: One who answers no would be thought to be unclean, but not the whole family.
  • 5. 326 How to do (or not to do) . . . Box 5. Testing the meaning of a ‘yes’ or ‘no’ response By undertaking a relatively simple interview we could quickly establish whether a ‘yes’ or ‘no’ response could be inter- preted clearly and as intended. This particular question was dropped because it became clear that we could not be sure that a ‘no’ meant worse ‘health’. Interviewer: Do you get enough information about current events in Kenya? Respondent: No. Interviewer: If someone answers ‘no’ to this question, is he/she in a better or worse position than someone who answers ‘yes? Respondent: The one who answers ‘yes’ is in a better position. Interviewer: Why is this? Respondent: Information about current events in Kenya can assist people in this area by helping them develop, such as knowing different methods of farming, knowing about disease and outbreaks and ways of combating them, and also improvement of educational facilities. Interviewer: What other reasons are there that someone could answer ‘yes’? Respondent: Those who own radios and are literate. People can get access to this information from radios, newspapers, meetings and friends. Interviewer: What other reasons are there that someone could answer ‘no’? Respondent: Those who don’t have radios and don’t read newspapers. Everyone doesn’t want to have access to infor- mation about Kenya. Those who do are mostly interested in politics and current affairs. Those who aren’t are the old, illiterate and those not interested in politics. itself in helping us learn about the questions. They were poor; and those with harmonious relationships compared selected purposively to include equal proportions of males with those in disputes. and females, as well as being roughly representative of the • As the questionnaire was intended to be able to measure different generations in the community. In total each ques- change we asked eight people for one of two types of hypo- tion was the focus of between four and eight interviews with thetical considerations of the questionnaire. For example, a range of people from the local population. members of a youth-based water and agriculture project were asked to imagine how they might have answered the questions before and after the project, and other key Targeted methods informants were asked how they might have completed Targeted methods focus on a particular component of the the questionnaire during and after malaria. The task was response process, to gain a detailed insight into the respon- followed up with a discussion of the answers most likely to dent’s comprehension and preparation of a response and change and why. to evaluate whether they performed the task the way the • To address which conception of health people felt the researcher intended. We used three types of targeted question represented, as well as helping further investi- methods: gations into the meaning and interpretation of the ques- tion, respondents were asked to sort the Kikamba • People with known (and opposite) characteristics of inter- questions into groups, either under the Kikamba words est were asked to complete specific questions to assess how used to translate the English word ‘health’7 that each ques- likely it was that the questionnaire would be interpreted in tion was most closely related to, or to devise their own the same way by people with particular characteristics. This categories and group them. Each question was placed on a was important in being able to interpret any future differ- separate card for ease of sorting. The task was followed up ences/similarities across groups. Therefore we contacted with an interview about the meanings of the Kikamba (through the help of our field assistants): people with and words used to translate the English word ‘health’ and without sickness/illness; the relatively wealthy and very reasons for groupings.
  • 6. How to do (or not to do) . . . 327 Box 6. Group discussion on ‘people in your homestead’ Many of the KENQOL questions ask respondents about people in their homestead,i for example, • Do people in your homestead listen to your advice/suggestions at the moment? • When you are sick, do people in your homestead help you in getting treatment? • At the moment is it easy for people in your homestead to find transport to go for treatment when they are sick? To check who people thought about when answering these questions, we provided several groups of respondents with an example family (see below) and initiated a discussion about who different people in that family would think about when they were asked specific questions. Example family: Headed by a married male (mzee) and female (mama), who had three sons. The first son is married to one wife with a child and they live just outside the compound. The second son is married with two wives (polygy- nous marriage) and both wives have children. The third son lives away in Nairobi. The discussion revealed that people would make reference to different people in their response. In the example above, the mzee would consider seven adults, whereas the first-born son thinks of his wife and his wife thinks only of her husband. To complicate matters further still, we discovered that a person could alter the group of people they thought about depending on the question that was being asked, even if each question used identical phrases for ‘people in your homestead’. We noted all the instances in which this occurred. i The phrase musyi kwaku/kwenyu for homestead was used after finding that the word family was interpreted very differently across people. Group methods questions and issues 1–6 in Table 1. The local field assistants also reviewed the written Kikamba questions and gave their Group methods also focus on a particular component of the views on issues 1–6 as well as 13–17. They were very helpful response process, but a group of people are addressed, rather in their criticisms and suggestions, and once we had been able than an individual, to encourage greater discussion around to take down their advice, it was also easier for them to inter- the issues involved. There were a number of key issues within view others (as opposed to discussing their own ideas with the survey questions that were repeated often, such as time interviewees). frames, the use of singular or plural ‘you’ and terms such as ‘neighbours’, ‘friends’ and ‘family’. Individual interviews were very helpful in understanding interpretations but did Step 4: Reviewing questions for inclusion, revising the not give us the opportunity to see opposing suggestions question or intended meaning, or dropping questions debated. We therefore held group discussions on focused The field transcripts and results from each research method issues. The group discussions lasted 1–2 hours, and were led were collated and summarized by the authors for each survey by two moderators. Each interview was recorded and later question. Reviews of questions were supported by a dis- transcribed. Each group involved people with similar ages cussion between the authors and field assistants. Following and gender, and the sum of groups captured the range of ages these discussions, decisions were made concerning the action and gender in the area. Box 6 gives an example of our required for each question including: approach to finding out how people interpreted the word ‘homestead’. • acceptance of the original question and meaning; • acceptance of the original question, with a change to the meaning; Expert evaluation • changing the question (slightly) but keeping the meaning Expert evaluation involved no interaction with potential the same; survey respondents, but rather those who could offer particu- • dropping the question; lar guidance on specific issues. The ‘experts’ included two • writing a new question (and new intended meaning). groups of people: external advisors including Kikamba Bible translators and Kamba university lecturers on sociology; and Any changes were tested through additional expanded the field assistants living locally in Makueni who had assisted interviews and any new questions also went through various the KENQOL research group for several years. The external translation checks. Prior to accepting any new or changed language advisors reviewed the written survey questions and questions all data on the question were once again collated focused particularly on their interpretation of the meaning of and reviewed prior to being taken forward to piloting.
  • 7. 328 How to do (or not to do) . . . Piloting into other languages. There has been some use of qualitative techniques to assess the content validity of original (e.g. Moser and Kalton (1992) refer to piloting as the ‘dress Saxena et al. 1998), but more usually the adapted versions of, rehearsal’. Questions are placed together as it is expected that generic HRQL instruments (e.g. Chwalow et al. 1992; Lam et they will appear on the final questionnaire and the dynamics of al. 1994), as well as the applicability of generic instruments to the survey as a whole are investigated. The KENQOL group specific disease groups (e.g. Tazaki et al. 1998). These investi- undertook piloting in two stages. The first focused on ques- gations tend to be relatively unstructured, and can be tions from each sub-section (n = 6) of the survey. The second described as cursory investigations. In general there is involved straight runs through the entire survey. One of the too much emphasis placed on quantitative assessments of authors was present at each of these interviews, and a field reliability, validity and sensitivity. The quantitative assess- assistant conducted the interview in Kikamba. At both stages ment is a rather closed approach, that ignores the respon- of piloting, the interviewer noted down the response to each dent’s own interpretation of questions. question (yes, no, don’t know, no response). They also noted any other information the respondent gave, such as asking for We therefore adopted a qualitative approach to the assess- clarification of the question, or the respondent providing ment of validity, reliability and sensitivity. This made both the additional information after responding ‘yes’, ‘no’ or ‘don’t researcher’s and the respondent’s interpretation of questions know’. In addition, the authors noted any body language, more open and accountable, and ensured that the final survey pauses before a response, emotional reactions to questions etc. questions were more reliable and valid and able to detect At the end of the pilot survey interviews, each respondent was change as revisions were made. Earlier we defined a reliable asked for their opinion about the questions, including: measure as one which yields the same results in repeated trials (Carmine and Zeller 1979). A question may be un- • what they thought about the questions in general; reliable due to bad wording, causing a person to understand • whether any of the questions seemed to be strange or the question differently on different occasions (De Vaus unusual; 1995). This paper has described procedures for assessing the • their opinion on the order of questions; reliability of questions during the pre-testing stage, for • how appropriate the response categories were; example identifying questions that cannot elicit a double • whether the correct procedures are used to address: negative for a response, do not contain high level or old fash- (a) married and unmarried people, ioned language, and are free from jargon. A valid measure is (b) old and young people; one which measures what is intended (Carmine and Zeller • whether any questions should not be asked in the survey; 1979), and methods described in this paper which helped to • to point out any questions they do not want to answer, or address this included ensuring that the questions capture think they should not be asked; current views, reflect local issues and the conception of health • any problems they thought we might encounter asking each question represents is clear. Similarly, during pre-testing people the questions: the sensitivity to questions over time was investigated. De (a) alone, Vaus (1995) explains that instead of this procedure some (b) with others listening (e.g. young people were asked people include a large number of indicators and at a later whether they might have inhibitions answering ques- stage eliminate those that do not seem to be working. tions in front of parents); However, he explains: ‘this seems to be the wrong way of • whether any questions appeared to be asking the same doing things since we will end up by defining the concept in thing; terms of the indicators which ‘worked’. If this is done then the • their opinion on what changes we could make to improve indicators may not represent the concepts of the theory we this survey in terms of: set out to test and as such the research can end up having (a) introduction, little relevance to the original research question’ (De Vaus (b) questions; 1995: 54). • which questions might be difficult for someone to talk about in front of an interviewer who comes from the same However, we recognize that the process was labour intensive area as the respondent; and took a considerable amount of time (around 6 months of • whether it matters who interviews people; researcher and 12 months of field assistant’s time). In our case • who would be a good interviewer in terms of age, gender we judged this to be an important part of the process and and any other characteristics. outcome of our research. First, we are intending that this measure becomes used across many different diseases as Based on feedback during piloting, several additional an outcome measure used to judge the effectiveness and revisions were made to the KENQOL questions. Each efficiency of health-improving interventions. It is therefore change was followed up by further piloting (and pre-testing important to us that the people whose health we are trying to where necessary) until the final version of the KENQOL measure (and improve) are integrated into this process, and survey was agreed. it is often the case that generic HRQL instruments are devel- oped and used for years following their development. Sec- ondly, it provided valuable comparative data for explaining Discussion any similarities and dissonances in measured health in Vihiga We have been concerned about the procedures used in the district, Kenya. Related to this is the possibility that others development of HRQL measures as well as their translations would want to translate the instrument for use elsewhere, and
  • 8. How to do (or not to do) . . . 329 we wanted to improve the basis from which judgements are required information through household survey, readers made about the equivalence of future translations. should consider some of the literature from demographic studies; some examples include Caldwell and Igun (1997), There are other techniques that are available to researchers Ewbank (1981), Bledsoe and Pison (1994). We found cogni- to test their survey questions that we did not use. The tive research assessment a particularly inspiring approach as KENQOL researchers adopted some cognitive assessment it captures approaches given to a variety of psychological survey methodologies, but there are others that may be more processes involved in information retrieval. Some of this appropriate for other research groups (Jobe and Mingay literature has already been cited above, but readers might 1989): also like to consider Lessler and Tourangeu (1989) and the proceedings from a seminar on the Cognitive Aspects of • Concurrent think-aloud interviews: respondents think Survey Methodology (Sirken et al. 1999). Researchers in the aloud when answering questions, and responses are field of marketing have also been considering these issues for probed. some time, consider for example Gendall and Hoek (1990) • Retrospective think-aloud interviews: respondents answer and Kumar et al. (1999). all question first and then are asked how they arrived at their answers. • Confidence ratings: respondents relate the degree of confi- Conclusions/recommendations dence they have in the accuracy of their answers. The investigations described in this paper should not be seen • Paraphrasing: respondents repeat the questions in their as an unusual step in the development and testing of survey own words. questionnaires – particularly those that will be used repeat- • Response latency: measurement of the elapsed time edly. The validity and reliability (and possibly sensitivity) of between the questions and respondent’s answers. any survey is important to establish prior to accepting results and conclusions. Each of these ideas can begin to be explored, The KENQOL group also adopted some methods recognized and improved, through the use of qualitative methods in pre- as participatory or rapid rural appraisal (PRA and RRA) testing and piloting questionnaires. Indeed, evidence that a approaches, for example the use of anthropologists and local survey tool has ‘content validity’ is required, rather than leaders as key informants (Chambers 1983). Other useful simply relying on unsubstantiated assertions by researchers, RRA and PRA approaches might include a well-being (or and only once this has been demonstrated is it appropriate to HRQL) ranking exercise of local households with local adopt quantitative assessment methods for further analysis. participants, and linking this to likely responses to questions Finally, the knowledge gained about the interpretation of and analysis of difference based on responses to questions survey questions will also help in judging the equivalence (see Chambers 1997). of any future translations of the survey questionnaire. Whilst the approach we have taken to the qualitative assess- ment of meaning attached to the KENQOL survey questions Endnotes is unique amongst generic HRQL research, many of the ideas 1 Validity determines that an instrument is measuring what was were drawn from a variety of sources, including survey intended. It concerns the crucial relationship between concept and research (e.g. Moser and Kalton 1992; Fowler 1993), trans- indicator. The objective is to ensure that the indicator represents the lation studies (e.g. Larson 1984; Barnwell 1992), the cognitive intended (and only the intended) concept (Carmines and Zeller sciences (e.g. Jobe 1990; Forsyth and Lessler 1991) as well as 1979). 2 Reliability concerns the extent to which an experiment, test the health measurement literature (e.g. Guillemin et al. 1993; and any measuring procedure yields the same results on repeated Herdman et al. 1997, 1998). De Vaus (1995) is particularly trials (Carmines and Zeller 1979). useful because of the links made between investigating ques- 3 The ability of a measure to detect change. tionnaire interpretation and the reliability and validity of a 4 Because most household interviews happen outside the home questionnaire. This gives an indication of the applicability of with many people around. To ask to interview people privately in a our approach to the design of survey questions more gener- household survey tends to arouse a lot of suspicion. 5 Our household survey eventually consisted of 112 health ally. As Tourangeau (1984) states ‘[in] the respondent’s task measurement questions and 50 socioeconomic and demographic . . . there is considerable room for error. Respondents may questions. We achieved a total of 550 interviews – a response rate of misunderstand the question or the response categories; they 81.3% and a mean item response rate of 98.1% (SD 3.6). may forget or misremember the crucial information; they may 6 All interviews were undertaken in Kikamba, the local lan- misjudge the information they do recall; and they may mis- guage of the Kamba population we were working amongst. The report their answer’ (Tourangeau 1984: 73–74). Kikamba versions of the KENQOL questions were used for all survey testing, although the English versions are reported here. The pre-test interviews were written up in full in English to facilitate Finally, to help researchers give greater attention to investi- participation by the non-Kikamba speaking researchers. gating whether respondents interpret questions in ways 7 wailu, wianie, utheu, useo, wiwa nesa, uima. researchers expect (or hope), we give a brief introduction to a variety of different types of source texts. Most social research texts address questionnaire design, but few provide References detail on investigations into question meanings. One text that Barnwell K. 1992. Bible Translation: An introductory course in trans- is useful in this context is Oppenheim (2000). For an excellent lation principles (3rd edition). Dallas, TX: Summer Institute of insight into the many pitfalls associated with eliciting the Linguistics International.
  • 9. 330 How to do (or not to do) . . . Bledsoe C, Pison G. 1994. Introduction. In: Bledsoe C, Pison G (eds). Lessler JT, Tourangeau R. 1989. Vital health statistics. Series 6.1. Nuptiality in Sub-Saharan Africa: Contemporary anthropo- Washington, DC: National Center for Health Statistics. logical and demographic perspectives. Oxford: Clarendon Press, Moser CA, Kalton G. 1992. Survey methods in social investigation pp. 1–22. (2nd edition). Aldershot, UK: Gower. Bowden A, Fox-Rushby J. 2000. A critical review of the adaptation Oppenheim AN. 2000. Questionnaire design, interviewing and and use of generic HRQL measures amongst populations attitude measurement. Buckingham, UK: Open University outside of North America, western/southern/northern Europe, Press. Australia and New Zealand. Quality of Life Research 9: 313. Saxena S, Chandiramani K, Bhargava R. 1998. WHOQOL-Hindi: A Caldwell JC, Igun AA. 1971. An experiment with census-type age questionnaire for assessing the quality of life in health care set- enumeration in Nigeria. Population Studies 25: 287–302. tings in India. National Medical Journal of India 11: 160–5. Carmines EG, Zeller RA. 1979. Reliability and validity assessment. Sirken M, Jabine T, Willis G, Martin E, Tucker C. 1999. A New London: Sage Publications. Agenda for Interdisciplinary Survey Research Methods: Pro- Chambers R. 1983. Rural development: putting the last first. Harlow, ceedings of the CASM II Seminar. Hyattsville, MD: National UK: Longman. Center for Health Statistics. [http://www.cdc.gov/nchs/data/ Chambers R. 1997. Whose reality counts? Putting the first last. casm2pro.pdf, 2nd August 2001] London: Intermediate Technology. Tazaki M, Nakane Y, Endo T et al. 1998. Results of a qualitative and Chwalow AJ, Lurie A, Bean K et al. 1992. A French version of the field study using the WHOQOL instrument for cancer patients. Sickness Impact Profile (SIP): stages in the cross-cultural vali- Japanese Journal of Clinical Oncology 28: 134–41. dation of a generic quality of life scale. Fundamental Clinical Tourangeau R. 1984. Cognitive sciences and survey methods. In: Pharmacology 6: 319–26. Jabine TB (ed). Cognitive aspects of survey methodology: build- De Vaus. 1995. Surveys in social research (4th edition). London: ing a bridge between disciplines. Washington, DC: National Routledge. Academy Press, pp. 73–100. Ewbank DC. 1981. Age misreporting and age-selective undernumer- ation: sources, patterns, and consequences for demographic analysis. Committee on Population and Demography. Report Acknowledgements No. 4. Washington, DC: National Academy Press. Forsyth BH, Lessler JT. 1991. Cognitive laboratory methods: a tax- We thank the following organizations for the funding of this survey: onomy. In: Biemer PP, Groves SM, Lyberg LE, Mathiowetz UNDP/World Bank/WHO Special Programme for Research and NA, Sudan S (eds). Measurement errors in surveys. New York: Training in Tropical Disease, the UK Economic and Social Research John Wiley, pp. 393–418. Council (Ref. H52427005994) and the Health Economics and Fowler FJ. 1993. Survey research methods (2nd edition). London: Financing Programme (funded by the Department for International Sage Publications. Development). We also thank all our field assistants, especially Fox-Rushby J, Parker M. 1995. Culture and the measurement of Justus Mutunga, Jeremiah Nyamai and Alex Muvea, for their help, health related quality of life. European Review of Applied Psy- guidance and enthusiasm. Finally we thank all the people who chology 45: 257–63. participated in the preparation for this survey and made us welcome Fox-Rushby J, Johnson K, Mwanzo I et al. 1997. Creating an instru- in their homes. ment to assess lay perceptions of HRQL: options and impli- cations. Quality of Life Research 6: 633. Fox-Rushby J, for the KENQOL Group. 2000. Operationalising con- Biographies ceptions of ‘health’ amongst the Wakamba and Maragoli of Annabel Bowden, Ph.D. (Health Geography), was a Research Fellow Kenya: the basis of the KENQOL instrument. Quality of Life at the London School of Hygiene and Tropical Medicine (LSHTM) Research 9: 316. at the time of this study. She was involved with the KENQOL project Gendall P, Hoek J. 1990. A question of wording. Marketing Bulletin for 5 years and was based in Kenya for 1 year during the household 1: 25–36. survey. She also has a wider interest in the applicability of Grosh M, Glewwe P. 2000. Designing household survey question- adapted/translated versions of generic health-related quality of life naires for developing countries: lessons from 15 years of the measures. She is now based with the Lewin Group (UK). Living Standard Measurement Survey (Volume 1). Washington, DC: World Bank. Julia Fox-Rushby, Ph.D. (Economics), is a Senior Lecturer at the Guillemin F, Bombardier C, Beaton D. 1993. Cross-cultural adap- LSHTM. She has written numerous academic papers on the cost- tation of health-related quality of life measures: literature effectiveness of health interventions across the world, specializing review and proposed guidelines. Journal of Clinical Epidemiol- particularly in maternal and child health, malaria and, more recently, ogy 46: 1417–32. vaccine preventable disease. She has also been involved over the past Herdman M, Fox-Rushby JA, Badia X. 1997. ‘Equivalence’ and the 15 years in developing a number of non-disease specific measures of translation and adaptation of health-related quality of life ques- health-related quality of life as a member of the EuroQol group, tionnaires. Quality of Life Research 6: 237–47. advisor to the WHOQOL group and as principal investigator of the Herdman M, Fox-Rushby JA, Badia X. 1998. A model of equival- KENQOL group. ence in the cultural adaptation of HRQL measures: a univer- salist approach. Quality of Life Research 7: 323–35. Jobe JB. 1990. Research on questionnaire design: perspectives from Lilian Nyandieka, MA (Sociology), is a Research Officer at the other disciplines. American Journal of Epidemiology 132: 824. Center for Public Health Research, a department of the Kenya Jobe JB, Mingay DJ. 1989. Cognitive research improves question- Medical Research Institute (PO Box 20752, Nairobi, Kenya). She has naires. American Journal Public Health 79: 1053–5. been involved in the KENQOL project for 3 years and is also Kenya Government. 2000. Makueni District Development Report researching social and economic aspects of malaria. 1997–2001. Nairobi: Office of the Vice President and Ministry of Planning and National Development. John Wanjau, BA (Anthropology), is an Assistant Research Officer Kumar V, Aaker DA, Day GS. 1999. Essentials of marketing at the Center for Public Health Research, Kenya Medical Research research. Chichester, UK: John Wiley. Institute. He has been involved in the KENQOL project for 3 years. Lam CL, Van Weel C, Lauder IJ. 1994. Can the Dartmouth COOP/ WONCA charts be used to assess the functional status of Correspondence: Julia Fox-Rushby Ph.D., Health Policy Unit, Chinese patients? Family Practice 11: 85–94. London School of Hygiene and Tropical Medicine, Keppel Street, Larson ML. 1984. Meaning based translation: a guide to cross-lan- London WC1E 7HT, UK. Tel: +44 (0) 20 7927 2267, Fax: +44 (0) 20 guage equivalence. Lanham, MD: University Press of America 7637 5391, Email: Julia.Fox-Rushby@lshtm.ac.uk