secondary analysis of qualitative data: a review of … · as the electronic databases did not...

32
SECONDARY ANALYSIS OF QUALITATIVE DATA: A REVIEW OF THE LITERATURE REF: R000222918 Janet Heaton Research Fellow Social Policy Research Unit University of York August 2000 • Heslington • York • YO10 5 DD

Upload: others

Post on 06-Sep-2019

2 views

Category:

Documents


0 download

TRANSCRIPT

SECONDARY ANALYSIS OFQUALITATIVE DATA: A REVIEW OF THELITERATURE

REF: R000222918

Janet HeatonResearch FellowSocial Policy Research UnitUniversity of York

August 2000

• Heslington • York • YO10 5 DD

SECONDARY ANALYSIS OFQUALITATIVE DATA: A REVIEW OF THELITERATURE

ESRC 1752 JH 8.00

Janet HeatonResearch FellowSocial Policy Research UnitUniversity of York

REF: R000222918

• Heslington • York • YO10 5 DD

ii

CONTENTS

Page

1. SUMMARY OF RESEARCH RESULTS 1

2. FULL REPORT OF RESEARCH ACTIVITIESAND RESULTS 4

Background 4Objectives 4Methods 5Results 6Activities 19Outputs 19Impacts 20Future research priorities 20

3. REFERENCES 21

4. APPENDIX: ‘SECONDARY ANALYSES OF QUALITATIVEDATA: AN ANNOTATED BIBLIOGRAPHY’ 26

1

SUMMARY OF RESEARCH RESULTS

Secondary analysis (SA) is a term used to describe various analytical practices which use pre-existing data to investigate new research questions or to re-examine primary study questionsfor purposes of corroboration. While the SA of quantitative data is an established and welldocumented methodology, re-use of qualitative data remains under-developed.

This study had three main aims:• to describe the nature and use of SA of qualitative data sets in the international health

and social care literature;• to critically appraise these studies, examining the epistemological, methodological and

ethical issues arising from the use of SA in this area of research;• to consider the overall quality and impact of the studies and implications for future

development of the methodology.

The nature and use of qualitative secondary analysis Fifty-five studies were identified which were self-defined as SA (N=36) or which fitted theabove definition (N=19). The vast majority of studies (N=49) were published in the 1990sand were by authors based in North America (N=47). Only seven studies originated from theUK. A large proportion of the studies (N=48) were authored by at least one person who wasinvolved with the primary research from which the data were derived. Thus, researcherswere tending to re-use their own data, sometimes sharing it informally with others, ratherthan drawing on independently collected and archived data sets.

Appraisal of these studies has revealed six types of SA being undertaken:

Supra analysis: transcends the focus of the primary studyfrom which the data were derived, examining new empirical,theoretical or methodological questions.

Amplified analysis: combines data from two or more primarystudies for purposes of comparison or enlarging sample.

Supplementary analysis: a more in-depth investigation of anemergent issue or aspect of the data which was not addressed inthe primary study.

Complementary analysis: the SA is supported by additionalprimary research or, alternatively, a primary study whichincludes an element of SA.

2

Alternative analysis: data are re-analysed using new methodsand/or perspectives for purposes of corroboration based on theprinciple of triangulation.

Repeat analysis: data are re-analysed using a similaranalytical framework in order to verify the findings of theprimary research.

Issues arising from the appraisal of studies Qualitative SA was found to differ from its quantitative counterpart in that it places emphasison the re-use of data sets to investigate new research questions (rather than the re-investigation of primary study questions for purposes of corroboration). This may reflect thedifferent epistemological foundations of research using qualitative and quantitative methods;thus, the latter methods are mainly used in the more positivistic social, natural and biomedicalsciences where greater emphasis is placed on the need to corroborate findings through re-analysis or re-studies.

The finding that qualitative SA was generally conducted by researchers who were involved inthe primary studies from which the data were derived not only further distinguishes the SA ofqualitative data from that of quantitative data but also raises issues about the relationship ofthe analyst to the data. Is the analyst who has direct knowledge and/or experience of thecontext of the data collection in a better position to conduct a SA? While this may be anadvantage, levels of involvement in primary research do vary. In addition, it may be possiblefor secondary analysts using archived data to consult with the primary researchers.Moreover, all analysts, whether or not they were ‘there’ at data collection, produce (primaryand secondary) analyses which are socially contingent.

Poor reporting of the conduct of the studies meant that it was not possible to assess theiroverall quality. Many failed to report on key issues, including the selection of the data set(s)and sub-sample(s), whether informed consent was sought from informants, how the ‘fit’ ofthe data and the focus of the SA was ascertained, how the rigour of the analysis wasestablished, and the possible limitations of the methodology. Where there was evidence thatresearchers had made efforts to establish the rigour of their analyses, it was not clear from thereports how the techniques employed, most of which are used in primary research, weretransferred and applied in the context of SA. However, there was evidence of methods, suchas grounded theory, being adapted specifically for use in this context.

Only two studies reported seeking informed consent from the informants whose data wasused for the SA. As the majority of studies failed to discuss the topic, it is not clear if consent

3

was simply presumed or if it was sought but not reported. This finding raises the issue ofwhether informed consent should be obtained in all secondary analyses and, if so, how. Arelated concern is whether the sharing of data violates prior agreements concerning theconfidentiality of the information. The evidence from this review suggests that it may beinappropriate to generalise about the need to obtain informed consent for secondary analyses,as this is likely to vary according to the characteristics of the secondary study.

Implications for future development of the methodology The following implications for the future development of the methodology were identified:• There is a need to better define the use of SA in reports and in electronic databases.• There is a need more fully to document the ways in which SA was accomplished in

reports of secondary studies.• The possibility of qualitative researchers extending their primary work through the SA

of auto-data or through informal data sharing should be acknowledged and facilitated.• Social science professional guidelines on the procurement of informed consent and

preservation of confidentiality need to be revised so that they consider these issuesmore specifically in the context of secondary research.

• Further research is required: - on the use of SA of qualitative data in other subject areas; - on researchers’ attitudes toward the use of archived qualitative data; - to establish what constitutes a worthy qualitative data set for archiving and re-use; - on ‘how to do’ SA of qualitative data; - on the public understanding of social science research and, in particular, lay views

on the re-use of data. This includes the circumstances in which informedconsent should be sought for secondary research, as well as the meaning ofconfidentiality and appropriateness of related techniques for anonymisingdata.

4

FULL REPORT OF RESEARCH ACTIVITIES AND RESULTS

BACKGROUND Secondary analysis (SA) is a term used to describe various analytical practices which makeuse of pre-existing data either to investigate new research questions or to re-examine primarystudy questions for purposes of corroboration.(1-13) Data used in this context are therebytransformed into ‘secondary data’. This data, which can be quantitative or qualitative or amixture of the two, may have been collected in the context of a research study (hereafterreferred to as ‘data sets’) or for other purposes (referred to as ‘naturalistic’ data, such asdiaries and administrative records). Although meta-analysis and literature reviews involvethe use of raw and/or published data from existing studies, these methodologies can bedistinguished from SA in that they are concerned with synthesising research findings, ratherthan examining new research questions or verifying results from individual studies.(4, 12)

This study is concerned with the SA of qualitative data sets. Unlike the SA of quantitativedata, which is an established and well documented methodology,(1, 3, 6, 7) the re-use ofqualitative data sets remains under-developed. Although the potential for re-using this typeof data was recognised as long ago as 1962,(14) SA has been mainly confined to statisticaldata from large scale social surveys and other sources. However, recent developments in theUK have both galvanised interest in the SA of qualitative data sets and made themethodology a more realistic proposition. In particular, the establishment of Qualidata by theEconomic and Social Research Council (ESRC) in 1994 and the adoption of policiespromoting qualitative data archiving and re-use by the ESRC and other research fundingorganisations has created the conditions of possibility for increased sharing of qualitativedata. There is therefore a need to examine existing efforts to pioneer the SA of qualitativedata sets and to consider how the methodology might best be developed in the future. Thisreview builds on the emerging literature on this topic, which has begun to outline therudiments of the methodology and its potential uses,(10-12) as well as the different forms it cantake.(4, 5)

OBJECTIVES The review had three main objectives:• to describe the nature and use of SA of qualitative data sets in the international health

and social care literature;• to critically appraise these studies, in particular examining the epistemological,

methodological and ethical issues arising from the use of SA in this area of research;

5

• to consider the overall quality and impact of the studies and the implications for futuredevelopment of the methodology.

The focus on the international health and social care literature was chosen for two mainreasons. First, it was already known that this was an area where SA of qualitative data hadbeen developed.(4) Secondly, the range of methods utilised, and the sensitive nature of someof the topics covered in this area, made it ideal for exploring the theoretical, technical andethical issues arising from the re-use of qualitative data sets. All the objectives wereaddressed, although the poor reporting of the methods used in the studies limited theassessment of the quality of this work and related issues, such as the possible influence ofnewly developed computer software for analysing qualitative data. The recency of many ofthe studies also restricted the extent to which their impact could be considered.

METHODS A number of resources were searched (electronically and manually) to identify secondaryanalyses of qualitative data: electronic databases (including BIDS Social Science CitationIndex, CINAHL, EMBASE, Medline, and Sociological Abstracts); e-journals; electronicsearch engines; publishers’ electronic bibliographies; key journals; study references; and keyauthor citations. Appeals for information were also issued via the British SociologicalAssociation Network magazine, conference flyers, electronic discussion lists (QUALRS-Land ARCHIVE-QUALITATIVE-DATA), and personal contacts.

As the electronic databases did not keyword ‘secondary analysis of qualitative data’ a numberof synonymous terms, such as ‘secondary interview data’ and ‘secondary-analysis withqualitative-studies’, were derived from the studies initially identified. These were usedsystematically to search study titles and abstracts for examples of relevant studies. Abstractsidentified by these means were read and the studies obtained if they:• employed SA of qualitative data sets (as defined above, p.3);• were in the area of health and social care;• were published in English.

Once obtained, the studies were checked for relevance and appraised. In some cases authorswere approached to clarify whether or not they defined the study as an example of SA.Details of each of the studies were entered into a spreadsheet (QuattroPro). An annotatedbibliography was also compiled using a reference management software package (EndNote)(see appendix).

6

While the review was wide-ranging, it cannot claim to be exhaustive, mainly because ofproblems with the definition of SA. As we will see, the review includes several studieswhich, although not self-defined as SA, nevertheless exemplify characteristics of themethodology. This finding confirms Thorne’s(12) observation that secondary analyses are notalways readily defined as such.

RESULTS The results of the review are presented under three main headings corresponding to theobjectives of the study.

Extent and nature of use of secondary analysis A total of 55 studies which were self-defined as SA (N=36) or which fitted the definition(N=19) were identified by the review. Other terms authors employed to describe the latterstudies included: ‘post hoc analysis’, ‘reanalysis’ and ‘retrospective latent content analysis’.The vast majority of studies (N=49) were published in the 1990s and were by authors basedin North America (N=47). Only seven studies originated from the UK. While allowing thatthese findings may reflect the geographical coverage of some of the resources searched andthe limitations of the search strategies adopted (which excluded studies not published inEnglish), they nevertheless suggest that the methodology is more developed in NorthAmerica than elsewhere.

Studies were classified into one of three groups depending on the relationship of the author(s)to the data. The largest group of studies (N=32) were by ‘non-independent’ authors -analysts who had been involved in the primary research and were therefore re-using their ownpreviously collected data (hereafter referred to as ‘auto-data’). In the second group of 16‘semi-independent’ studies one or more - but not all - of the authors had been involved in theprimary study. Thus, 48 out of the 55 studies were authored by at least one person who wasinvolved with the primary research. Of the remaining studies, six were classified as‘independent’ secondary analyses as the author(s) had not been involved with the primaryresearch, and one study used both auto - and independent data sets.

The majority of studies (N=40) were secondary analyses of a single data set. Twelve studiesdrew on two data sets and three utilised more than two data sets in total. While secondaryface-to-face interview data were utilised in 46 studies, qualitative data derived fromobservational work, focus groups, surveys, telephone interviews, vignettes, documents,autobiographies and published ethnographies were also drawn upon. In half the secondaryanalyses a subset of the data from the primary study was selected for re-use; the other half

7

utilised the full data set. A small number of studies (N=5) also involved the collection ofadditional primary or secondary (naturalistic) data.

8

Previous overviews of the methodology have drawn attention to various possible types of SAof qualitative data.(4, 5, 11) Building on this work, the studies identified in this first extensiveempirical review of secondary analyses were examined for variations in their approach. Thestudies were found to vary on five main dimensions:• function - whether or not the SA was designed to investigate new research questions or

to re-investigate a primary study question for purposes of corroboration;• focus - how the question(s) addressed by studies investigating new research questions

differed from those addressed in the primary research;• source(s) of data - whether one or more secondary data sets were used and whether any

primary research was also conducted alongside the SA;• methods/perspectives - whether these were similar to or different from the approach

employed in the primary research;• origins of secondary data - the extent to which studies relied on auto-data and/or data

from other sources.

Based on this information, six varieties of SA were discerned: supra analysis, amplifiedanalysis, supplementary analysis, complementary analysis, alternative analysis and repeatanalysis. While some of these types of SA are similar to those previously identified, othersare new. Hence, this typology refines and extends previous models of classification,providing knowledge of the range of approaches to the SA of qualitative data in the area ofhealth and social care research.

Following an adumbration of the typology in Table 1, each of the types are fully describedbelow together with some illustrative examples of relevant studies.

Table 1: Typology of secondary analysis of qualitative data

Supra analysis Transcends the focus of the primary study from which the data werederived, examining new empirical, theoretical or methodologicalquestions.

Amplified analysis Combines data from two or more studies for purposes of comparison orenlarging sample.

Supplementary analysis A more in-depth investigation of an emergent issue or aspect of the datawhich was not addressed in the primary study.

Complementaryanalysis

The SA is supported by additional primary research; or a primary studywhich includes an element of SA.

Alternative analysis Data is re-analysed using new methods and/or perspectives for purposesof corroboration based on the principle of triangulation.

9

Repeat analysis Data is re-analysed using a similar analytical framework in order tocheck the findings of the primary research.

10

It is important to note that these are ideal types and hence individual studies may not alwaysexhibit all the characteristics of a given type and, indeed, may share attributes of differenttypes. Accordingly, studies have been classified as exemplifying a particular type (or types)of SA on the basis of their principal characteristics. A summary of the number and keycharacteristics of the different varieties of SA are set out in tabular form in Table 2 at the endof this section.

Supra analysis ‘Supra analyses’ transcend or go beyond the terms of the primary or ‘parent’ study fromwhich the data were derived. Such studies involve the investigation of new empirical,theoretical or methodological questions. An equivalent approach was previously highlighted(but unnamed) by Heaton(4) and shares some characteristics with a variety of SA called‘armchair induction’ identified by Thorne,(11) although supra analysis may be conducted bythe same researchers who carried out the primary research and is not necessarily restricted totheoreticians.

Eight out of the 55 studies developed analyses which transcended the terms of the primarywork, focussing on new aspects of the data and often employing a new theoreticalperspective.(15-22) Although not self-defined as SA, a study by Bloor and McIntosh(15)

usefully exemplifies the characteristics of supra analysis (as well as amplified analysis - seebelow). They analysed data from two separate studies of health visiting and therapeuticcommunities. The focus of their analysis was clearly not planned at the outset of the study:

‘Of course, the health visiting and therapeutic communities studies were notundertaken in order to compare professional-client relationships in different typesof service-provision, nor were they undertaken to elaborate a Foucauldianapproach to client resistance. The possibility of writing this chapter only occurredto us post hoc, when we realised that we both had data bearing upon issues ofpower and contest which showed both similarities and dissimilarities intechniques of client resistance.’ (p.161).

In focussing on forms of surveillance in professional-client relationships and associatedstrategies of resistance, the authors go beyond the terms of the original studies and develop anew, free-standing analysis consistent with supra analysis as it is defined here.

Amplified analysis This approach draws on two or more existing research data sets in order to examine commonand/or divergent themes across them. It can involve the comparison of different studypopulations or the pooling of data on a similar population. This definition incorporatesvariations of this approach previously described as ‘amplified sampling’,(11) ‘aggregated

11

analysis’,(23) and ‘pooled case comparison’,(24) which all involve the SA of multiple pre-existing data sets. Amplified analysis was the second most common type of SA identified by the review.Seventeen examples were found.(15, 16, 25-39) Of these, 12 made use of two data sets. All theamplified analyses were by authors who had been involved with the primary studies, althoughadditional independent researchers also collaborated with some of these secondary analyses.

Two main groups of amplified analysis were discerned. In the first, researchers whoseparately conducted independent primary studies later performed a SA of the data sets tofurther explore common issues across the study populations. For example, Yamashito andForsyth(39) learned, through meeting at a conference, that they had each conducted studies offamilies’ reactions to a relatives’ mental illness in Canada and the USA respectively. Thismeeting led to a SA in which they draw on ‘aggregated analysis’(23) to examine similaritiesand differences in the two data sets. In the other group of amplified analyses, researchers made use of two or more data sets fromtheir own oeuvre. This approach is illustrated by a series of studies by Sally Thorne andCarole Robinson. They combined the data from their MA projects and published twosecondary studies on health care relationships.(33, 38) Building on this work, theysubsequently obtained funding to do a primary study in 1989.(40) Individually the authorsthen produced four secondary analyses of these data sets (co-classified here as examples ofamplified/supplementary analyses). Thus, Thorne used the data to examine mothers’experiences of chronic illness and the meaning of non-compliance for those with chronicillness, as well as completing a doctoral thesis involving the analysis of secondary andprimary sources.(35-37) Robinson later drew on part of the same material to examine thenormalisation of life by chronically ill members and their families.(32) Although this laststudy is not self-defined as SA, it is set in the context of the aforementioned work and thetheme it explores was analysed separately and retrospectively. These examples of the re-analysis of data from one’s own oeuvre reflect an incremental approach to research wherebyindividual studies evolve progressively, rather than forming a disconnected series of works.

Supplementary analysis ‘Supplementary analysis’ involves a more in-depth focus on an emergent issue or aspect ofthe data which was not addressed (or was only partially addressed) by the primary analysis.Thus, the focus of the SA may shift to a particular theme, or to issues pertaining to a sub-sample of the primary study population. Supplementary analysis is in these respects relatedto Thorne’s ‘retrospective interpretation’ and ‘analytic expansion’ categories (the formerreferring specifically to researchers’ re-use of their own data).(12) Other examples of thisvariety of SA have been previously described by Heaton(4) and Hinds et al.(5)

12

As the foci of supplementary analyses are compatible with that of the primary study, the twomay be difficult to separate, particularly as researchers are increasingly selectively reportingthe results of primary research in multiple publications rather than in a single, overarchingpublication. However, while selective reporting may focus on a particular aspect of theprimary study, it is still produced in the context of the ongoing primary research. Bycontrast, supplementary analysis exceeds the analytical remit of the primary study, examiningin more depth a theme or sub set of the data which has emerged as a post hoc matter ofinterest and subsequent focus of inquiry.

This approach was found to be by far the most common form of SA of qualitative data,exemplified by 36 of the 55 studies identified.(29, 32, 33, 35-37, 41-70) In the majority of thesupplementary analyses (N=22) a subset of the data from the primary study was examined.For example, in one of two secondary analyses of data from their primary study of womencocaine users, Kearney et al. focussed in more depth on the mothers in this sample.(69) Theremainder of the supplementary analyses (N=14) involved the re-examination of an entiredata set. For instance, following on from one of the authors’ exploratory, descriptive study ofpain management at home by cancer patients, Vallerand and Ferrell’s secondary study on theconcept of control involved the use of the full data set.(68)

Complementary analysis This form of SA is combined with primary data collection and analysis. Such a study mayrely mainly on secondary data, with primary research conducted in support of the former.Equivalent examples of this approach were first identified by Hinds et al.(5) Alternatively, theanalysis of secondary data may form a relatively minor part of a primary study. This use ofSA has been defined as ‘cross-validation’ by Thorne.(11) Complementary analysis cantherefore be used to provide additional comparative or collateral evidence using differentsources of data, or to verify the results of the main analysis.

Four out of the five examples of complementary analyses identified were basically secondarystudies which involved some additional primary data collection and analysis.(37, 44, 48, 49) Twoof these studies - neither self-defined as secondary analyses - were by Hutchinson and madeuse of data from a study of unprofessional behaviour among hospital-based nurses.(48, 49) Inthe first ‘descriptive’ study, the complementary/supplementary analysis focussed on nurses’self-care strategies. A sub-set of the data was re-used and an additional 20 interviews wereconducted with nurses to augment the data set and to check initial observations. In thesecond ‘grounded theory’ study, the complementary/supplementary analysis focussed on how

13

nurses bend the rules for the sake of the patient. Again a sub-set of the data was re-used andan additional 21 interviews were conducted in order to ‘verify’ the initial observations.

14

Only one example of a primary study incorporating analysis of secondary data sources wasfound. Sandelowski and Jones undertook a primary study of couples’ experiences of pre-natal diagnosis and, in particular, their responses to the acquisition of foreknowledge of fetalimpairment.(71) Additional information from research on parenting medically fragile infantsby another investigative team was also used ‘for purposes of further comparison andvalidation’ (p.85).

Alternative analysis Unlike the preceding four types of SA which are all concerned with the investigation of newresearch questions, ‘alternative analysis’ maintains a focus on the question(s) posed in theprimary study but re-investigates these using new methods and/or analytical frameworks.This approach may therefore produce alternative findings or theoretical insights, or it mayconfirm and validate the original findings through triangulation. Alternative analysis has notbeen previously identified in overviews of the SA of qualitative data.

Only one example of this form of SA was found.(72) Re-using data from research onovereating among women who weight-cycle, the study aimed:

‘to analyse further subjects’ interview responses to ensure that no importantinformation was omitted in the primary analysis, which used a reversal theorycoding system only. The secondary analysis provided a validity check for theprimary coding results and an accuracy check for complete interpretation. Usingmethodological triangulation... two methods of viewing the same empiricalcontent, two coders performed a content analysis of the interview data with noconsideration for reversal theory’ (p.71).

Thus, in this case, the aims of the primary and secondary study were compatible, but themethods of analysis were different.

Repeat analysis Like alternative analysis, ‘repeat analysis’ maintains a focus on the primary study question(s)but instead uses a similar analytical approach with the intention of corroborating the findingsof the original research. Repeat analysis may in principle be conducted either by independentanalysts or by the original researchers in order to verify their findings. This approachepitomises the spirit of open scientific enquiry which has underpinned the development of thedata sharing movement in the USA.(73, 74)

While this type of SA is generally recognised in theory, no examples of it being applied inpractice to qualitative data were identified by this review. However, an example of a repeatanalysis of this type of data from another area is Lewis’s(75) re-analysis and re-study ofRedfield’s(76) research on life in a Mexican village.

15

Table 2: Principal characteristics of types of secondary analysis

TYPE & No.* FUNCTION FOCUS SOURCE(S) OF DATA METHODS/ PERSPECTIVES

ORIGINS OFSECONDARY DATA

SUPRAANALYSIS (N=8)

To investigate a newresearch question

Transcends primarystudy (PS)

Secondary data set(s) ormixed secondary andprimary data

Discontinuous orcontinuous with PS

Independent data orauto-data

AMPLIFIEDANALYSIS (N=17)

To investigate a newresearch question

Transcends PS or in-depth analysis of anaspect of the PSs

Multiple secondary data sets Continuous ordiscontinuous with PS

Independent data and/orauto-data

SUPPLEMENTARY ANALYSIS(N=36)

To investigate a newresearch question

In-depth analysis ofan aspect of the PS

Secondary data set(s) ormixed secondary andprimary data

Continuous ordiscontinuous with PS

Auto-data orindependent data

COMPLEMENTARY ANALYSIS(N=5)

To investigate a newresearch question orvalidate an ongoingprimary study (PS)

In-depth analysis ofan aspect of the PSor transcends PS

Mixed secondary andprimary data sets

Continuous ordiscontinuous with PS

Independent or auto-data

ALTERNATIVEANALYSIS (N=1)

To re-investigate (PS)question

Same focus Secondary data set Discontinuous withPS

Independent or auto-data

REPEATANALYSIS (N=0)

To re-investigate PSquestion

Same focus Secondary data set; may be partof a wider primary re-study

Continuous with PS Independent or auto-data

*The numbers add up to more than 55 because some studies shared the principal characteristics of more than one type.

16

17

Issues arising from the appraisal of studies Various epistemological, methodological and ethical issues were raised by the criticalappraisal of the studies. These will be discussed under separate headings, although someissues are clearly inter-related.

Epistemological issues To date, the concept of SA has mainly been defined in relation to the re-use of quantitativedata. Most commentators agree that it involves the use of pre-existing data either toinvestigate new research questions or to re-investigate primary study questions using newmethods or alternative perspectives. However, as we have seen, the secondary analysesidentified by this review have generally been concerned with the former. This may reflect thedifferent epistemological foundations of research using qualitative and quantitative-basedmethods; thus, the latter methods are mainly used in the more positivistic social, natural andbiomedical sciences where greater emphasis is placed on the need to corroborate findings orreplicate studies.

SA is also generally associated with the use of statistical data sets which were originallycollected by others(1, 3, 7, 9) although, as Hyman(6) observes, in the USA researchers did re-usetheir own survey data before data archives were established. While no one since Hyman hasconsidered this possibility in relation to quantitative data, in the emerging literature on theSA of qualitative data it has been recognised that the methodology can in principle involveauto-data as well as independently collected data sets.(4, 10-12)

As we have seen, auto-data was used in 48 of the 55 studies identified by the review. In re-using their own data, researchers potentially overcome one of the main objections to the SAof qualitative data - the problem of interpreting data originally collected by others.(4,5) Whilefamiliarity with the context of data collection is an advantage, levels of familiarity can vary.For example, the data may have been collected by a team of researchers with splitresponsibilities for the study design, data collection, processing and analysis.(4) Similarly, inthe case of amplified analysis, where two or more data sets from independent secondarysources are combined and jointly analysed, each researcher concerned will have only partialknowledge of these data. Secondary analyses may also be co-conducted with researcherswho were not involved in the primary study. In addition, as Mauthner et al. have pointed out,researchers’ ability to interpret their own data may also decline over time as memories wane;changes in researchers’ personal situation and new knowledge that they have gained since theprimary study may also influence their re-interpretation of the data.(77)

18

Only seven studies were found to use independent data sources. Of these, just three usedarchived material (and two of these used the same data set). Thus, Paget(57) used archiveddata from Waitzen’s(78) research and Weaver(21) developed a ‘secondary ethnographic dataanalysis’ of Roth’s data derived from his study of tuberculosis sanitoriums.(79) The latter dataset was also used by Weaver and Atkinson to illustrate a book on the use of computersoftware to assist qualitative data analysis.(22) Of the remaining studies, one was ofnaturalistic data in the form of lay autobiographies.(34) In another case, the author was‘invited’ by the primary researchers to conduct the SA.(42) And in yet another case the dataappeared to be obtained informally from another research team.(71) Finally, one SA(20) drewon a published analysis by Zussman,(80) as opposed to the raw data itself.

There are several possible reasons why archived data has not thus far been utilised to agreater extent. Archives for qualitative data have not been established as long as those forquantitative data and the availability of data sets, although increasing, is relatively limited atpresent. In addition, there may be advantages to sharing data on a more informal basis, as insome of the amplified analyses identified, whereby the SA is performed by a team ofresearchers who can pool their respective knowledge of the primary studies from which thedata were derived. The data concerned are also more easily shared on this basis, as comparedwith the relatively restricted access to data deposited in data archives and the associated costsinvolved.(81) Unlike quantitative data which tends to be shared electronically and inprocessed form, qualitative data sets comprised of tapes and transcripts are often available forconsultation in the archives only, hence it is less convenient to access this material anddevelop computer-assisted analyses. Finally, while the development of archives has begun toprovide material for researchers to re-use, nevertheless, the methodology of SA of qualitativedata has remained largely under-theorised, and hence researchers have lacked understandingof appropriate conceptual and methodological tools.

Methodological issues In general, the appraisal of the conduct and therefore the quality of the studies was hamperedby a lack of reporting on how these were actually accomplished. One of the best documentedsecondary analyses was by Szabo and Strang.(65) Their report provides information on: therespective funding of the primary and secondary work; the relationship of each of the authorsto the data; how informed consent and ethical approval for the SA was obtained; thecomposition of the data set; the use of grounded theory in the analysis; how the SA wasperformed; how the data were managed; and how the rigour of the analysis was established.However, many studies failed to report on key issues pertaining to the conduct of SA,including the selection of the data set(s) and sub-sample(s), how the ‘fit’ of the data and the

19

focus of the SA was ascertained, how the rigour of the analysis was established, and thepossible limitations of the methodology. In some studies, there was evidence that researchers had made efforts to establish the rigourof their analyses. The techniques employed included the establishment of audit trails,validation of findings using members of the study population, triangulation of data sources,consultation with the primary investigator(s), and independent coding of data. However, itwas not clear from the reports how these techniques, most of which are used in primaryresearch, were successfully transferred and applied in the context of SA.

Various methods, including grounded theory, narrative and content analysis, were applied inthe studies. Grounded theory was the most popular approach, used in some form or other byseveral studies. In certain cases, this approach had been adapted specifically for use in thecontext of SA. Thus, one aspect of grounded theory - theoretical sampling - whereby thestudy sample and/or research topics are revised in order to purposively address emergentthemes from the initial analysis, was deemed not possible because of the post hoc nature ofthe secondary studies. Hence, an alternative approach was developed, whereby theoreticalsampling was applied within the available sample in order to compare and contrast theexperiences of a given population.(56, 65, 69, 70) In addition, some complementary analyseseffectively overcame the limitations of working with pre-existing data by conductingadditional primary research to purposively address questions related to the focus of the SA.(48,

49)

Ethical issues Only two studies reported seeking informed consent from the informants whose data wasused for the SA. In one case this was obtained retrospectively(42) and in the otherprospectively.(65) As the vast majority of studies failed to discuss the topic, it is not clear ifconsent was simply presumed or if it was sought but not reported. This finding raises theissue of whether informed consent should be obtained in all secondary analyses and, if so,how. A related concern is whether the sharing of data violates prior agreements concerningthe confidentiality of the information. Existing professional codes of practice in the socialsciences vary in the attention given to these issues in the context of SA.(82-88) While theygenerally endorse data sharing in appropriate forms and stress the importance of obtaininginformed consent, at the same time, it is recognised that there are some situations, such ascovert research, where this requirement is waived. Researchers are, however, required toconsider possible harm to subjects and, according to the American Sociological Association(ASA),(82) to seek approval of institutional review boards (ethical committees) or relevantexperts.

20

The evidence from this review suggests that it may be inappropriate to generalise about theneed to obtain informed consent for secondary analyses, as this is likely to vary according tothe characteristics of the study. For example, in the case of supplementary analyses by a loneresearcher using auto-data, the purpose of this secondary study may be sufficiently related tothe primary study to be covered by the terms under which consent was originally obtained;furthermore, in re-using her own data, the researcher is not breaking confidentiality.Conversely, it is doubtful whether consent could be presumed in the case of supra analysisusing archived data.

Where informed consent for SA is considered necessary, obtaining this retrospectively maynot be feasible. Research participants’ names may be unavailable because of confidentialityagreements in the primary study; in addition, subjects may have moved or died since theywere involved in the primary research. Alternatively, consent may be obtained prospectivelyby primary researchers who prepare subjects for the possibility that the data may be stored inarchives and/or re-used for other purposes. While this is the more practical option, there aretwo main problems with it. The first is anticipating and explaining to informants exactly howthe data may be used in the future. Secondly, even where conditions are imposed on accessto and use of data by others, control over usage of the information cannot be fully guaranteed.

Implications for the development of the methodology Based on the findings of this review, the following nine key implications for the futuredevelopment of the methodology have been identified. These concern the conceptualisationof SA (i), the reporting of SA (ii), the promotion of SA (iii-iv), and areas where furtherresearch is required (v-ix).

i)There is a need to better define the use of SA in reports and in electronic databases. This review has confirmed Thorne’s observation that secondary analyses are not alwaysclearly defined as such.(12) The future classification and sub-classification of studies may behelped by this review which has shown how the methodology differs from primary analysis(in the form of selective reporting of research results), literature reviews and meta-analysis, aswell as distinguishing six types of SA of qualitative data.

ii) There is a need more fully to document the ways in which SA was accomplished inreports of secondary studies.

This includes the basic accounting of how data set(s) and any sub-sample(s) were selected,whether or not consent was sought and obtained, how the fit of the secondary researchquestion and the data were ascertained, how the rigour of the analysis was established, and

21

the limitations of the methodology. Reviewers and journal editors need to encourage authorsto more fully report on the conduct of secondary qualitative studies.

iii) The possibility of qualitative researchers’ extending their primary work through the SAof auto-data or through informal data sharing should be acknowledged and facilitated.

The main advantages of re-using these sources of data are convenience, researchers’familiarity with the context of the data collection, and that researchers maintain ownershipand control of the data.

iv) Social science professional guidelines on the procurement of informed consent andpreservation of confidentiality need to be revised so that they consider these issuesmore specifically in the context of secondary research.

The guidelines of the ASA are presently the most comprehensive in this respect.(82) In theUK, Qualidata, in conjunction with the ESRC, have also produced draft guidelines on theissues of confidentiality and copyright with respect to the collection of data and its possiblere-use.(85-88) These may form a useful basis for updating relevant professional guidelines.

v) Further research is required on the use of SA of qualitative data in other subject areas. This review has focussed solely on the international health and social care literature. Otherareas of research, such as education and crime, may provide evidence which will modifyand/or clarify some of the conclusions arrived at in this study. Such reviews might alsousefully consider the impact and utility of the methodology in social research, as it isdeveloped over time and applied more extensively.

vi) Further research is required on researchers’ attitudes toward the use of archivedqualitative data.

The lack of utilisation of this source of data raises issues about whether this is due to theavailability and adequacy of current stock, practical difficulties in accessing and analysing thematerial, or principled resistance to the formal sharing of qualitative data.

vii) Further research is required to establish what constitutes a worthy qualitative data setfor archiving and re-use.

This may be informed by related ongoing work to develop criteria for assessing the quality ofqualitative research (led by Professor J Popay and funded by the Health TechnologyAssessment programme).

viii) Further research is required on ‘how to do’ SA of qualitative data.

22

Future consideration of the methodology could usefully examine the respective use of auto -and independent data sets (from formal and informal sources), as well as the differentapproaches to SA outlined in this review. It could also consider techniques for establishingrigour, and the use of methods such as grounded theory in the context of SA.

ix) Finally, there is a need for future research on the public understanding of socialscience research and, in particular, lay views on the re-use of data. This includes thecircumstances in which informed consent should be sought for secondary research, aswell as the meaning of confidentiality and appropriateness of related techniques foranonymising data.

Such research could usefully explore the claims of Qualidata(89) that: ‘..most people do believe that research is for the public good and that theircontribution will be used in some way to create a better informed society, andeven go some way towards implementing policy changes’ (p.22).

In summary, this review has found that the SA of qualitative data has not yet been developedextensively but that different approaches have been pioneered which have potential for theconduct of original research using pre-existing data. In clarifying the conceptualisation of SAin relation to qualitative data and in highlighting areas for future research, it is hoped that thisreview will be instrumental in the future realisation of the potential of the methodology.

ACTIVITIES The following activities have been undertaken or are arranged:• ‘Secondary analysis of qualitative data.’ Poster presentation at the Social Policy

Association conference: Futures of Social Policy and Practice? University of Surrey atRoehampton, 18-20 July, 2000.

• ‘Secondary analysis of qualitative data.’ Poster presentation at the BSA MedicalSociology and European Society of Health and Medical Sociology joint conference:Health in Transition: European Perspectives. University of York, 14-17 September2000.

• ‘The possibility of secondary analysis’ Invited seminar presentation at theEpidemiology Department and Leicestershire Health Authority, University of Leicester.November 2000.

• Member of the Qualitative Methods Network, co-ordinated by Professor Jennie Popay,with proposed links to the Cochrane Methods Group and Campbell Collaboration.

OUTPUTS• ‘Secondary analysis of qualitative data: An annotated bibliography’. See Appendix.

This will be made available on the web.

23

• Contracted to publish a book based on the review with Sage (due December 2001).

IMPACTS Appeals for information on the Internet for references to secondary studies also generatedresponses from researchers interested in the project, including researchers, students andarchivists in the UK and abroad. Some studies which had not been self-defined as ‘secondaryanalyses’ in the publication were referred to me by the authors as possible examples. Otherresearchers whom I have contacted about their work have re-considered whether or not theirwork is best defined as ‘secondary analysis’. These initial responses suggest that the work,particularly when published in book form, will hopefully impact on how people conceptualiseand define such studies in the future.

My involvement in the Qualitative Methods Network is primarily because of the implicationsof this work for: distinguishing secondary analysis from related methodologies and meta-analysis in particular; reporting the methods adopted in this form of qualitative research; andfor developing methods for assessing the quality of qualitative work. While the Network isstill in the process of defining its remit and links with other groups, it is likely that this willhave a major international impact on the use of qualitative methods.

FUTURE RESEARCH PRIORITIES See Results, points 5-9 (p.24-25).

24

REFERENCES 1. Dale, A., Arber, S. and Proctor, M. (1988) Doing Secondary Analysis, London: UnwinHyman. 2. Glass, G.V. (1976) ‘Primary, secondary, and meta-analysis of research’. EducationalResearcher, 5(11): 3-8. 3. Hakim, C. (1982) Secondary Analysis in Social Research: A Guide to Data Sources andMethods with Examples, London: George Allen Unwin. 4. Heaton, J. (1998) ‘Secondary analysis of qualitative data’. Social Research Update,(22). 5. Hinds, P.S., Vogel, R.J. and Clarke-Steffen, L. (1997) ‘The possibilities and pitfalls ofdoing a secondary analysis of a qualitative data set’. Qualitative Health Research, 7(3): 408-424. 6. Hyman, H.H. (1972) Secondary Analysis of Sample Surveys: Principles, Procedures, andPotentialities, New York: John Wiley & Sons, Inc. 7. Kiecolt, K.J. and Nathan, L.E. (1985) Secondary Analysis of Survey Data, London: Sage. 8. McArt, E.W. and McDougal, L.W. (1985) ‘Secondary data analysis - a new approach tonursing research’. Image: The Journal of Nursing Scholarship, XVII(2): 54-57. 9. Miller, J.D. (1982) ‘Secondary analysis and science education research’. Journal ofResearch in Science Teaching, 19(9): 719-725. 10. Szabo, V. and Strang, V.R. (1997) ‘Secondary analysis of qualitative data’. Advances inNursing Science, 20(2): 66-74. 11. Thorne, S. (1994) ‘Secondary analysis in qualitative research: issues and implications’.In J. M. Morse (ed.) Critical Issues in Qualitative Research Methods, London: Sage. 12. Thorne, S.E. (1998) ‘Ethical and representational issues in qualitative secondaryanalysis’. Qualitative Health Research, 8(4): 547-555. 13. Woods, N.F. (1988) ‘Using existing data sources: primary and secondary analysis’. InN.F. Woods and M. Catanzaro (eds) Nursing Research: Theory and Practice, St. Louis: TheC.V. Mosby Company. 14. Glaser, B.G. (1962) ‘Secondary analysis: a strategy for the use of knowledge fromresearch elsewhere’. Social Problems, 10(1): 70-74. 15. Bloor, M. and McIntosh, J. (1990) ‘Surveillance and concealment: a comparison oftechniques of client resistance in therapeutic communities and health visiting’. In S.Cunningham-Burley and N.P. McKeganey (eds) Readings in Medical Sociology, London:Tavistock/ Routledge.

25

16. Fendrich, M., Wislar, J.S., Mackesy-Amiti, M.E. and Goldstein, P. (1996)‘Mechanisms of noncompletion in ethnographic research on drugs: results from a secondaryanalysis’. Journal of Drug Issues, 26(1): 23-44. 17. Heaton, J. ‘Hospital discharge and the temporal regulation of bodies’. Time & Society. 18. Jairath, N. (1999) ‘Myocardial infarction patients’ use of metaphors to share meaningand communicate underlying frames of experience’. Journal of Advanced Nursing, 29(2):283-289. 19. Jenny, J. and Logan, J. (1996) ‘Caring and comfort metaphors used by patients incritical care’. Image: Journal of Nursing Scholarship, 28(4): 349-352. 20. Pascalev, A. (1996) ‘Images of death and dying in the intensive care unit’. Journal ofMedical Humanities, 17(4): 219-236. 21. Weaver, A. (1994) ‘Deconstructing dirt and disease: the case of TB’. In M. Bloor and P.Taraborrell (eds) Qualitative Studies in Health and Medicine, Aldershot: Avebury. 22. Weaver, A. and Atkinson, P. (1994) Microcomputing and Qualitative Data Analysis,Aldershot: Avebury. 23. Estabrooks, C.A., Field, P.A. and Morse, J.M. (1994) ‘Aggregating qualitativefindings: an approach to theory development’. Qualitative Health Research, 4(4): 503-511. 24. West, J. and Oldfather, P. (1995) ‘Pooled case comparison: an innovation for cross-casestudy’. Qualitative Inquiry, 1(4): 452-464. 25. Angst, D.B. and Deatrick, J.A. (1996) ‘Involvement in health care decisions: parentsand children with chronic illness’. Journal of Family Nursing, 2(2): 174-194. 26. Arksey, H. and Sloper, P. (1999) ‘Disputed diagnoses: the cases of RSI and childhoodcancer’. Social Science & Medicine, 49: 483-497. 27. Backett, K.C. and Davison, C. (1995) ‘Lifecourse and lifestyle: the social and culturallocation of health behaviours’. Social Science and Medicine, 40(5): 629-638. 28. Bull, M.J. and Kane, R.L. (1996) ‘Gaps in discharge planning’. The Journal of AppliedGerontology, 15(4): 486-500. 29. Cohen, M.H. (1995) ‘The triggers of heightened parental uncertainty in chronic, life-threatening childhood illness’. Qualitative Health Research, 5(1): 63-77. 30. Ferrell, B.R., Grant, M., Dean, G.E., Funk, B. and Ly, J. (1996) ‘‘Bone tired’: theexperience of fatigue and its impact on quality of life’. Oncology Nursing Forum, 23(10):1539-1547. 31. Kirschbaum, M.S. and Knafl, K. (1996) ‘Major themes in parent-provider relationships:a comparison of life-threatening and chronic illness experience’. Journal of Family Nursing,2(2): 195-216.

26

32. Robinson, C.A. (1993) ‘Managing life with a chronic condition: the story ofnormalisation’. Qualitative Health Research, 3(1): 6-28. 33. Robinson, C.A. and Thorne, S. (1984) ‘Strengthening family ‘interference’’. Journal ofAdvanced Nursing, 9: 597-602. 34. Thorne, S.E. (1988) ‘Helpful and unhelpful communications in cancer care: the patientperspective’. Oncology Nursing Forum, 15(2): 167-172. 35. Thorne, S.E. (1990) ‘Mothers with chronic illness: a predicament of social construction’.Health Care for Women International, 11: 209-221. 36. Thorne, S.E. (1990) ‘Navigating troubled waters: chronic illness experience in a healthcare crisis’ The Union Institute of Advanced Studies, Cincinnati. 37. Thorne, S.E. (1990) ‘Constructive noncompliance in chronic illness’. Holistic NursingPractice, 5(1): 62-69. 38. Thorne, S.E. and Robinson, C.A. (1988) ‘Health care relationships: the chronic illnessperspective’. Research in Nursing & Health, 11: 293-300. 39. Yamashita, M. and Forsyth, D.M. (1998) ‘Family coping with mental illness: anaggregate from two studies, Canada and United States’. Journal of the American PsychiatricAssociation, 4(1): 1-8. 40. Thorne, S.E. and Robinson, C.A. (1989) ‘Guarded alliance: health care relationships inchronic illness’. Image: Journal of Nursing Scholarship, 21(3): 153-157. 41. Breckenridge, D.M. (1997) ‘Decisions regarding dialysis treatment modality: a holisticperspective’. Holistic Nursing Practice, 12(1): 54-61. 42. Clarke-Steffen, L. (1998) ‘The meaning of peak and nadir experiences of pediatriconcology nurses: secondary analysis’. Journal of Pediatric Oncology Nursing, 15(1): 25-33. 43. Deatrick, J.A., Knafl, K.A. and Guyer, K. ‘The meaning of caregiving behaviours:inductive approaches to family theory development’. In S. L. Feetham, S. B. Meister, J. M.Bell and C. L. Gilliss (eds) The Nursing of Families: Theory/Research/Education/Practice,Newbury Park: Sage. 44. Docherty, W.J., Lester, M.E. and Leigh, G. (1986) ‘Marriage encounter weekends:couples who win and couples who lose’. Journal of Marital and Family Therapy, 12(1): 49-61. 45. Gallo, A.M. and Knafl, K.A. (1998) ‘Parents’ reports of “tricks of the trade” formanaging childhood chronic illness’. Journal of the Society of Paediatric Nurses, 3(3): 93-102. 46. Gregory, D. and Longman, A. (1992) ‘Mothers’ suffering: sons who died of AIDS’.Qualitative Health Research, 2(3): 334-357.

27

47. Hall, J.M., Stevens, P.E. and Meleis, A.I. (1992) ‘Developing the construct of roleintegration: a narrative analysis of women clerical workers’ daily lives’. Research in Nursing& Health, 15: 447-457. 48. Hutchinson, S. (1987) ‘Self-care and job stress’. Image: Journal of Nursing Scholarship,19(4): 192-196. 49. Hutchinson, S.A. (1990) ‘Responsible subversion: a study of rule-bending amongnurses’. Scholarly Inquiry for Nursing Practice: An International Journal, 4(1): 3-17. 50. Jones, J.B. (1997) ‘Representations of menopause and their health care implications: aqualitative study’. American Journal of Preventive Medicine, 13(1): 58-65. 51. Katz, A. (1997) ‘’Mom, I have something to tell you’ - disclosing HIV infection’.Journal of Advanced Nursing, 25: 139-143. 52. Kidd, P., Scharf, T. and Veazie, M. (1996) ‘Linking stress and injury in the farmingenvironment: a secondary analysis of qualitative data’. Health Education Quarterly, 23(2):224-237. 53. Knafl, K.A., Ayres, L., Gallo, A.M., Zoeller, L.H. and Breitmayer, B. J. (1995)‘Learning from stories: parents’ accounts of the pathways to diagnosis’. Pediatric Nursing,21(5): 411-415. 54. Logan, J. and Jenny, J. (1990) ‘Deriving a new nursing diagnosis through qualitativeresearch: dysfunctional ventilatory weaning response’. Nursing Diagnosis, 1(1): 37-43. 55. McLaughlin, E. and Ritchie, J. (1994) ‘Legacies of caring: the experiences andcircumstances of ex-carers’. Health & Social Care, 2: 241-253. 56. Messias, D.K.H., Yeager, K.A., Dibble, S.L. and Dodd, M.J. (1997) ‘Patients’perspectives of fatigue while undergoing chemotherapy’. Oncology Nursing Forum, 24(1):43-48. 57. Paget, M.A. (1983) ‘On the work of talk: studies in misunderstandings’. In S. Fisher andA. D. Todd (eds) The Social Organisation of Doctor-Patient Communication, Washington:The Center for Applied Linguistics. 58. Powers, B.A. (1996) ‘Relationships among older women living in a nursing home’.Journal of Women & Aging, 8(3-4): 179-198. 59. Princeton, J.C. (1993) ‘Education for executive nurse administrators: a databasedcurricular model for doctoral (PhD) programs’. Journal of Nursing Education, 32(2): 59-63. 60. Rehm, R.S. and Catanzaro, M.L. (1998) ‘‘It’s just a fact of life’: family members’perceptions of parental chronic illness’. Journal of Family Nursing, 4(1): 21-40. 61. Rush, K.L. and Ouellet, L.L. (1997) ‘Mobility aids and the elderly client’. Journal ofGerontological Nursing, 23(1): 7-15.

28

62. Sandelowski, M. (1994) ‘Channel of desire: fetal ultrasonography in two use-contexts’.Qualitative Health Research, 4(3): 262-280. 63. Sandelowski, M. (1994) ‘Separate, but less unequal: fetal ultrasonography and thetransformation of expectant mother/fatherhood’. Gender & Society, 8(2): 230-245. 64. Sandelowski, M. and Black, B.P. (1994) ‘The epistemology of expectant parenthood’.Western Journal of Nursing Research, 16(6): 601-622. 65. Szabo, V. and Strang, V.R. (1999) ‘Experiencing control in caregiving’. Image: Journalof Nursing Scholarship, 31(1): 71-75. 66. Thorne, S.E., Harris, S.R., Hislop, T.G. and Vestrup, J.A. (1999) ‘The experience ofwaiting for diagnosis after an abnormal mammogram’. The Breast Journal, 5(1): 42-51. 67. Tishelman, C. and Sachs, L. (1998) ‘The diagnostic process and the boundaries ofnormality’. Qualitative Health Research, 8(1): 48-60. 68. Vallerand, A.H. and Ferrell, B.R. (1995) ‘Issues of control in patients with cancerpain’. Western Journal of Nursing Research, 17(5): 467-483. 69. Kearney, M.H., Murphy, S. and Rosenbaum, M. (1994) ‘Mothering on crack cocaine:a grounded theory analysis’. Social Science & Medicine, 38(2): 351-361. 70. Kearney, M.H., Murphy, S. and Rosenbaum, M. (1994) ‘Learning by losing: sex andfertility on crack cocaine’. Qualitative Health Research, 4(2): 142-162. 71. Sandelowski, M. and Jones, L.C. (1996) ‘Couples’ evaluations of foreknowledge offetal impairment’. Clinical Nursing Research, 5(1): 81-96. 72. Popkess-Vawter, S., Brandau, C. and Straub, J. (1998) ‘Triggers of overeating andrelated intervention strategies for women who weight cycle’. Applied Nursing Research,11(2): 69-76. 73. Fienberg, S.E., Martin, M.E. and Staff, M.L. (1985) Sharing Research Data,Washington DC: National Academy Press. 74. Sieber, J.E. (ed) (1991) Sharing Social Science Data: Advantages and Challenges,London: Sage. 75. Lewis, O. (1972) Life in a Mexican Village: Tepoztlan Restudied, Urbana: The Universityof Illinois Press. 76. Redfield, R. (1964) Tepoztlan - A Mexican Village: A Study of Folk Life, Chicago: TheUniversity of Chicago Press. 77. Mauthner, N.S., Parry, O. and Backett-Milburn, K. (1998) ‘The data are out there, orare they? Implications for archiving and revisiting qualitative data’. Sociology, 32(4): 733-745.

29

78. Waitzkin, H. (1991) The Politics of Medical Encounters: How Patients and DoctorsDeal with Social Problems, London: Yale University Press. 79. Roth, J. (1963) Timetables: Structuring the Passage of Time in Hospital Treatment andOther Careers, Indianapolis: Bobbs-Merrill. 80. Zussman, R. (1992) Intensive Care: Medical Ethics and the Medical Profession,Chicago: University of Chicago Press. 81. Hammersley, M. (1997) ‘Qualitative data archiving: some reflections on its prospectsand problems’. Sociology, 31(1): 131-142. 82. American Sociological Association (ASA) ‘Ethical Standards’. Available on web(August 2000) at http://www.asanet.org/members/ecostand.html 83. Social Research Association (SRA) ‘Ethical Guidelines’. Available on web (August2000) at http://www.the-SRA.org.uk/ethics.htm 84. British Sociological Association (BSA) ‘Statement of Ethical Practice’. Available onweb (August 2000) at http://www.britsoc.org.uk/ethgu2.htm 85. Economic and Social Research Council (ESRC) (1998) ‘Your views upon copyrightand confidentiality’. Draft guidelines available on web (August 2000) athttp://www.escr.ac.uk/ guide.htm 86. ESRC Qualitative Data Archival Resource Centre (QUALIDATA) ‘Guidelines onconfidentiality and informed consent’. Available on web (August 2000) at http://www.essex.ac.uk/qualidata/forms/confiden.htm 87. ESRC Qualitative Data Archival Resource Centre (QUALIDATA) ‘Guidelines oncopyright’. Available on web (August 2000) athttp://www.essex.ac.uk/qualidata/forms/copyright .htm 88. ESRC Qualitative Data Archival Resource Centre (QUALIDATA) ‘Confidentiality,consent and copyright in the interviewing of children’. Available on web (August 2000) athttp://www.essex.ac.uk/qualidata/forms/children.htm 89. Corti, L. (1999) ‘Text, sound and videotape: The future of qualitative data in the globalnetwork’. IASSIST Quarterly, 23(2): 18-25.