Health Policy and Planning, 35, 2020, 1424–1431

doi: 10.1093/heapol/czaa064
Advance Access Publication Date: 11 November 2020
Methodological Musings

Document analysis in health policy research:

the READ approach
1,2, 3 1,4
Sarah L Dalglish *, Hina Khalid and Shannon A McMahon

1
Department of International Health, Johns Hopkins School of Public Health, 615 N. Wolfe St, Baltimore, MD 21205,

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

USA 2Institute for Global Health, University College London, Institute for Global Health 3rd floor, 30 Guilford Street,
London WC1N 1EH, UK 3School of Humanities and Social Sciences, Information Technology University, Arfa
Software Technology Park, Ferozepur Road, Lahore 54000, Pakistan and 4Heidelberg Institute of Global Health,
Medical Faculty and University Hospital, University of Heidelberg, Im Neuenheimer Feld 130/3, 69120 Heidelberg,
Germany
*Corresponding author. Department of International Health, Johns Hopkins School of Public Health, Baltimore, MD, USA.
E-mail: sdalgli1@jhu.edu
Accepted on 17 May 2020

Abstract
Document analysis is one of the most commonly used and powerful methods in health policy
research. While existing qualitative research manuals offer direction for conducting document ana-
lysis, there has been little specific discussion about how to use this method to understand and ana-
lyse health policy. Drawing on guidance from other disciplines and our own research experience,
we present a systematic approach for document analysis in health policy research called the READ
approach: (1) ready your materials, (2) extract data, (3) analyse data and (4) distil your findings. We
provide practical advice on each step, with consideration of epistemological and theoretical issues
such as the socially constructed nature of documents and their role in modern bureaucracies. We
provide examples of document analysis from two case studies from our work in Pakistan and
Niger in which documents provided critical insight and advanced empirical and theoretical under-
standing of a health policy issue. Coding tools for each case study are included as Supplementary
Files to inspire and guide future research. These case studies illustrate the value of rigorous docu-
ment analysis to understand policy content and processes and discourse around policy, in ways
that are either not possible using other methods, or greatly enrich other methods such as in-depth
interviews and observation. Given the central nature of documents to health policy research and
importance of reading them critically, the READ approach provides practical guidance on gaining
the most out of documents and ensuring rigour in document analysis.

Keywords: Health policy, health systems research, interdisciplinary, methods, policy, policy analysis, policy research, qualitative,
research methods, social sciences

Introduction
Document analysis (also called document review) is one of the most analysed official reports, laws, statues, census reports and newspapers
commonly used methods in health policy research; it is nearly impos- and periodicals over a nearly 50-year period to come to his world-
sible to conduct policy research without it. Writing in early 20th cen- altering conclusions (Harvey, 1990). Yet in much of social science re-
tury, Weber (2015) identified the importance of formal, written search, ‘documents are placed at the margins of consideration,’ with
documents as a key characteristic of the bureaucracies by which mod- privilege given to the spoken word via methods such as interviews,
ern societies function, including in public health. Accordingly, critical possibly due to the fact that many qualitative methods were developed
social research has a long tradition of documentary review: Marx in the anthropological tradition to study mainly pre-literate societies

C The Author(s) 2020. Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine.
V
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits
unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. 1424
Health Policy and Planning, 2020, Vol. 35, No. 10 1425

Key Messages

• Rigour in qualitative research is judged partly by the use of deliberate, systematic procedures; however, little specific
guidance is available for analysing documents, a nonetheless common method in health policy research.
• Document analysis is useful for understanding policy content across time and geographies, documenting processes, tri-
angulating with interviews and other sources of data, understanding how information and ideas are presented formally,
and understanding issue framing, among other purposes.
• The READ (Ready materials, Extract data, Analyse data, Distil) approach provides a step-by-step guide to conducting
document analysis for qualitative policy research.
• The READ approach can be adapted to different purposes and types of research, two examples of which are presented
in this article, with sample tools in the Supplementary Materials.

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

(Prior, 2003). To date, little specific guidance is available to help document analysis and provide two short case studies demonstrating
health policy researchers make the most of these wells of information. how document analysis can be applied.
The term ‘documents’ is defined here broadly, following Prior,
as physical or virtual artefacts designed by creators, for users, to
function within a particular setting (Prior, 2003). Documents exist What is document analysis?
not as standalone objects of study but must be understood in the Document analysis is a systematic procedure for reviewing or evalu-
social web of meaning within which they are produced and con- ating documents, which can be used to provide context, generate
sumed. For example, some analysts distinguish between public questions, supplement other types of research data, track change
documents (produced in the context of public sector activities), pri- over time and corroborate other sources (Bowen, 2009). In one com-
vate documents (from business and civil society) and personal monly cited approach in social research, Bowen recommends first
documents (created by or for individuals, and generally not meant skimming the documents to get an overview, then reading to identify
for public consumption) (Mogalakwe, 2009). Documents can be relevant categories of analysis for the overall set of documents and
used in a number of ways throughout the research process (Bowen, finally interpreting the body of documents (Bowen, 2009).
2009). In the planning or study design phase, they can be used to Document analysis can include both quantitative and qualitative
gather background information and help refine the research ques- components: the approach presented here can be used with either set
tion. Documents can also be used to spark ideas for disseminating of methods, but we emphasize qualitative ones, which are more
research once it is complete, by observing the ways those who will adapted to the socially constructed meaning-making inherent to col-
use the research speak to and communicate ideas with one laborative exercises such as policymaking.
another. The study of documents as a research method is common to a
Documents can also be used during data collection and analysis number of social science disciplines—yet in many of these fields,
to help answer research questions. Recent health policy research including sociology (Mogalakwe, 2009), anthropology (Prior,
shows that this can be done in at least four ways. Frequently, pol- 2003) and political science (Wesley, 2010), document-based re-
icy documents are reviewed to describe the content or categorize search is described as ill-considered and underutilized.
the approaches to specific health problems in existing policies, as Unsurprisingly, textual analysis is perhaps most developed in fields
in reviews of the composition of drowning prevention resources in such as media studies, cultural studies and literary theory, all disci-
the United States or policy responses to foetal alcohol spectrum plines that recognize documents as ‘social facts’ that are created,
disorder in South Africa (Katchmarchi et al., 2018; Adebiyi et al., consumed, shared and utilized in socially organized ways
2019). In other cases, non-policy documents are used to examine (Atkinson and Coffey, 1997). Documents exist within social ‘fields
the implementation of health policies in real-world settings, as in a of action,’ a term used to designate the environments within which
review of web sources and newspapers analysing the functioning individuals and groups interact. Documents are therefore not mere
of community health councils in New Zealand (Gurung et al., records of social life, but integral parts of it—and indeed can be-
2020). Perhaps less frequently, document analysis is used to ana- come agents in their own right (Prior, 2003). Powerful entities also
lyse policy processes, as in an assessment of multi-sectoral plan- manipulate the nature and content of knowledge; therefore, gaps
ning process for nutrition in Burkina Faso (Ouedraogo et al., in available information must be understood as reflecting and po-
2020). Finally, and most broadly, document analysis can be used tentially reinforcing societal power relations (Bryman and Burgess,
to inform new policies, as in one study that assessed cigarette sticks 1994).
as communication and branding ‘documents,’ to suggest avenues Document analysis, like any research method, can be subject to
for further regulation and tobacco control activities (Smith et al., concerns regarding validity, reliability, authenticity, motivated
2017). authorship, lack of representativity and so on. However, these can
This practice paper provides an overarching method for conduct- be mitigated or avoided using standard techniques to enhance quali-
ing document analysis, which can be adapted to a multitude of re- tative rigour, such as triangulation (within documents and across
search questions and topics. Document analysis is used in most or methods and theoretical perspectives), ensuring adequate sample
all policy studies; the aim of this article is to provide a systematized size or ‘engagement’ with the documents, member checking, peer
method that will enhance procedural rigour. We provide an over- debriefing and so on (Maxwell, 2005).
view of document analysis, drawing on guidance from disciplines Document analysis can be used as a standalone method, e.g. to
adjacent to public health, introduce the ‘READ’ approach to analyse the contents of specific types of policy as they evolve over
1426 Health Policy and Planning, 2020, Vol. 35, No. 10

time and differ across geographies, but document analysis can also The READ approach
be powerfully combined with other types of methods to cross-
The READ approach to document analysis is a systematic procedure
validate (i.e. triangulate) and deepen the value of concurrent meth-
for collecting documents and gaining information from them in the
ods. As one guide to public policy research puts it, ‘almost all likely
context of health policy studies at any level (global, national, local,
sources of information, data, and ideas fall into two general types:
etc.). The steps consist of: (1) ready your materials, (2) extract data,
documents and people’ (Bardach and Patashnik, 2015). Thus,
(3) analyse data and (4) distil your findings. We describe each of
researchers can ask interviewees to address questions that arise from
these steps in turn.
policy documents and point the way to useful new documents.
Bardach and Patashnik suggest alternating between documents and
interviews as sources as information, as one tends to lead to the Step 1. Ready your materials
other, such as by scanning interviewees’ bookshelves and papers for At the outset, researchers must set parameters in terms of the nature
titles and author names (Bardach and Patashnik, 2015). Depending and number (approximately) of documents they plan to analyse,
on your research questions, document analysis can be used in com- based on the research question. How much time will you allocate to
bination with different types of interviews (Berner-Rodoreda et al., the document analysis, and what is the scope of your research ques-

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

2018), observation (Harvey, 2018), and quantitative analyses, tion? Depending on the answers to these questions, criteria should
among other common methods in policy research. be established around (1) the topic (a particular policy, programme,

Panel 1 Exploring the framing of diseases in Pakistani media

Health policies must account for how societies perceive and understand a given disease’s origins and causes, and media sources play an important
role in framing health issues (Nelkin, 1991; Entman, 1993). Document analysis was employed to understand the frames used in print media (newspa-
pers) in Pakistan when discussing Human Immunodeficiency Virus (HIV) and viral hepatitis, two diseases that are spread using similar modes of trans-
mission but have varying levels of stigma in the country. Alongside document analysis, key informant interviews were used for triangulation and to
flesh out what stigma for HIV meant in the country.
A sample of newspaper articles was drawn from the electronic database LexisNexis (January 2006-September 2016) based on readership, electronic
availability in LexisNexis and geographic diversity, to capture cultural differences across provinces over time (Strömbäck and Dimitrova, 2011). Broad
search terms were used for HIV and viral hepatitis, resulting in 3415 articles for hepatitis and1580 articles for HIV. A random sample comprising 10% of
the total HIV articles (n = 156) and 5% of the total hepatitis articles (n = 176) was selected and coded using a fixed coding guide. The coding guide was
developed using an inductive approach (Krippendorff, 2004; Mayring, 2004), which involved reading a sample of articles line by line to identify media
frames for HIV and viral hepatitis (Abdelmutti and Hoffman-Goetz, 2009; Claassen et al., 2012). Two rounds of pre-testing were carried out before the
final sample of articles was coded. However, the use of LexisNexis as the primary data source excluded newspapers published in the local language (open-
ing up the possibility of omitting some media frames). Therefore, interviews were important for triangulation of findings.
Data from document analysis were collated in an Excel sheet and analysed in STATA 14. The findings of the document analysis highlighted that while
both diseases were transmitted predominantly through injecting drug use in the country, hepatitis was only discussed using frames such as ‘medical’ (dis-
cussing transmission, prevention, and treatment methods), ‘resources’ (resources available to fight the disease), ‘magnitude’ (gives the scope of the problem
or disease prevalence) and ‘need for awareness’–there was no ‘stigma and discrimination’ frame attached to the disease [Figure, HIV and viral hepatitis
articles by main frames (%)]. In contrast, the ‘stigma and discrimination’ frame and the ‘social causes of disease’ frame (discussing non-medical causes for
the spread of disease) were used exclusively in articles on HIV, notably including suggestions that acquiring the disease was linked to socially immoral and
un-Islamic behaviour. Key informant interviews helped to probe further the traits associated with someone who had HIV. Taken together, document ana-
lysis and key informant interviews helped build a richer narrative of HIV stigma in the country.
Given the difference in how these diseases were understood, these findings suggested that there was a need for explicit policy to reframe HIV as a disease.
Countries such as Iran, Indonesia and Malaysia have successfully garnered government and policy attention to HIV and reduced stigma by reframing it as
a disease spread through injecting drug use (Kamarulzaman, 2013).

HIV and viral hepatitis articles by main frames (%). Note: The percentage of articles is calculated by dividing the number of articles appearing in each frame for
viral hepatitis and HIV by the respectivenumber of sampled articles for each disease (N = 137 for HIV; N = 117 for hepatitis). Time frame: 1 January 2006 to 30
September 2016
Health Policy and Planning, 2020, Vol. 35, No. 10 1427

Table 1 Types of documents that can be consulted in studies of health policy

Category Examples

Official documents • Policies or policy directives

• Strategies for sectors or on specific health problems
• Official statements and declarations
• Official position papers
• Statistical surveys or publications
Implementation documents • Training manuals or work tools (booklets, clinical files, etc.)
• Midterm/final reports or evaluations
• Financial analyses
• Operational plans
• Project proposals
• Funding requests
Legal documents • Laws

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

• Regulations
• Memorandums of understanding
• Cooperation agreements
Working documents • Meeting reports or minutes
• Memoranda
• Committee reports
• PowerPoint presentations
• Draft documents
• Mission reports
• Emails
Scholarly work • Scientific or peer-reviewed publications
• Masters or doctoral dissertations
• Textbooks and other course materials
Media and communications • Newspaper and magazine articles
• Podcasts, videos and radio and television segments
• Advertisements and posters
• Newsletters, bulletins, listservs, blogs and webpages
• Twitter conversations and other social media
Other • Promotional materials (pens, notebooks, lanyards, etc.)
• Warning labels and nutritional labels on food and other products
• Medical or other health devices
• Floor plans, architectural plans and maps

or health issue, narrowly defined according to the research ques- questions or consider other methods of inquiry. If you have too
tion); (2) dates of inclusion (whether taking the long view of several many documents, you can either analyse a subset of these (Panel 1)
decades, or zooming in on a specific event or period in time); and (3) or adopt more stringent inclusion criteria.
an indicative list of places to search for documents (possibilities in- In Table 1, we present a non-exhaustive list of the types of docu-
clude databases such as Ministry archives; LexisNexis or other data- ments that can be included in document analyses of health policy
bases; online searches; and particularly interview subjects). For issues. In most cases, this will mean written sources (policies, reports,
difficult-to-obtain working documents or otherwise non-public articles). The types of documents to be analysed will vary by study
items, bringing a flash drive to interviews is one of the best ways to and according to the research question, although in many cases, it
gain access to valuable documents. will be useful to consult a mix of formal documents (such as official
For research focusing on a single policy or programme, you may policies, laws or strategies), ‘gray literature’ (organizational materials
review only a handful of documents. However, if you are looking at such as reports, evaluations and white papers produced outside for-
multiple policies, health issues, or contexts, or reviewing shorter mal publication channels) and, whenever possible, informal or work-
documents (such as newspaper articles), you may look at hundreds, ing documents (such as meeting notes, PowerPoint presentations and
or even thousands of documents. When considering the number of memoranda). These latter in particular can provide rich veins of in-
documents you will analyse, you should make notes on the type of sight into how policy actors are thinking through the issues under
information you plan to extract from documents—i.e. what it is you study, particularly for the lucky researcher who obtains working
hope to learn, and how this will help answer your research ques- documents with ‘Track Changes.’ How you prioritize documents will
tion(s). The initial criteria—and the data you seek to extract from depend on your research question: you may prioritize official policy
documents—will likely evolve over the course of the research, as it documents if you are studying policy content, or you may prioritize
becomes clear whether they will yield too few documents and infor- informal documents if you are studying policy process.
mation (a rare outcome), far too many documents and too much in- During this initial preparatory phase, we also recommend devis-
formation (a much more common outcome) or documents that fail ing a file-naming system for your documents (e.g.
to address the research question; however, it is important to have a Author.Date.Topic.Institution.PDF), so that documents can be eas-
starting point to guide the search. If you find that the documents ily retrieved throughout the research process. After extracting data
you need are unavailable, you may need to reassess your research and processing your documents the first time around, you will likely
1428 Health Policy and Planning, 2020, Vol. 35, No. 10

have additional ‘questions’ to ask your documents and need to con- source files and link filenames to the data that you are extracting
sult them again. For this reason, it is important to clearly name (see sample naming conventions in the Supplementary Materials).

Panel 2 Case study Documents tell part of the story in Niger

In a multi-country policy analysis of integrated Community Case Management of childhood illness (iCCM), Niger was among the few countries that
scaled up the policy at national level (Bennett et al., 2015). Alongside key stakeholder interviews and non-participant observation, document analysis
was used to reconstruct the policy process leading to this outcome.
In total, 103 documents were obtained from policy actors in Niger, researchers working on similar topics, or collected on the Internet (Dalglish et
al., 2015). Documents included official policies and strategies, field reports, legal regulations, program evaluations, funding proposals, newsletters and
newspaper articles, among other sources. Document acquisition was greatly facilitated by asking for documents during stakeholder interviews, al-
though some documents were not available due to a fire that destroyed World Health Organization (WHO) servers in the years preceding the study.
Data from the documents was extracted into a Microsoft Excel file, recording information about specific aspects of child health policy and programs,
framing of issues, use of research evidence, and mention of international recommendations, among other topics. Documents were also used to compile
a timeline of events in the policy process.

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

Policy processes were elucidated by creating a timeline of events, which documented how specific decrees, workshops, meetings, and other events
occurred over time. The timeline was overlaid with measures of implementation (number of health posts built, number of health workers trained) to
understand how decision-making processes propelled real-world outcomes, and served as proxies for financial data that were rarely included in policy
documents (Dalglish et al., 2015).
Additionally, document analysis revealed a partial account of what was driving these events. Many documents showed a concern for reaching the
Millennium Development Goal on child mortality (Figure, Representations of progress toward Millennium Development Goal 4 in Nigerien policy
documents). Graphs mapping country progress toward Millennium Development Goal (MDG)-4 appeared in nearly all documentation on iCCM, and
progress was regularly reported on by the Nigerien National Institute of Statistics, suggesting that these were a significant motivating factor in policy
and resource allocation decisions. Yet older historical documents showed a long-standing recognition of the problem of children’s access to life-saving
healthcare (well before the MDGs), with policy remedies going back to least 1965 in the form of rural first-aid workers (Fournier and Djermakoye,
1975). Triangulation with interviews and observation also showed that national policymakers’ practical knowledge and ethical imperative to save
children’s lives was at least as important as the MDGs in motivating policy action (Dalglish et al., 2017). Taken together, the document and non-docu-
ment data showed that, as in other contexts, the MDGs were useful mainly to direct international fundraising and satisfy donor norms in expectation
of funding increases (Marten, 2019).

Representations of progress toward Millennium Development Goal 4 in Nigerien policy documents. Sources: clockwise from upper left: (WHO 2006); (Institut
National de la Statistique 2010); (Ministè re de la Santé Publique 2010); (Unicef 2010)
Health Policy and Planning, 2020, Vol. 35, No. 10 1429

Table 2 Questions to ask your overall body of documents

When analysing individual documents:

• Is the document complete? Is this a finished document, or a draft?
• What is the purpose of the document? Who is its target audience?
• Under which circumstances was the document produced? Under which circumstances is it consumed?
• Who created the document? Aside from the listed authors, what other contributors were likely involved in its creation?
• What could be the ‘agenda’ of the document’s creators?
• Are there other versions of the document? Why? How might they differ?
• Are there internal contradictions within the document (e.g. differing rationales or framings)?
• Is the document credible? Do you have any questions about its accuracy, good faith, balance, selective reasoning, etc.?
• What sources are cited (or not cited)? What kind of evidence does it use?
When analysing the overall body of documents:
• How complete is the set of documents? What is missing?

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

• Which documents were easy to find? Which were harder to find? Which proved impossible to find? Why might this be the case?
• Which voices are represented in the overall body of documents? Which are not?
• How do the documents compare in terms of content? How do they compare in terms of style, format, length and ‘look’? How about in terms of
formality and tone?
• What visual information can you find in the documents (charts and graphs, pictures, etc.)?
• How are the same issues discussed in different ways across documents?
• Do the documents ‘speak to each other’? Do they reference each other, or respond to each other’s arguments or propositions? Are they responding
to other documents not included in the review?
• How are documents similar or different across different topic areas, types of document or governance levels (e.g. global, national, sub-national)?
• How does the information from documents compare to data from other data sources (e.g. interviews, focus groups, observation, quantitative
analyses)?

Step 2. Extract data you have already read and processed, which will be greatly facili-
Data can be extracted in a number of ways, and the method you se- tated by a coherent file-naming system. It is also useful to keep notes
lect for doing so will depend on your research question and the na- on other documents that are mentioned that should be tracked
ture of your documents. One simple way is to use an Excel down (sometimes you can write the author for help). As a general
spreadsheet where each row is a document and each column is a cat- rule, we suggest being parsimonious when selecting initial categories
egory of information you are seeking to extract, from more basic to extract from data. Simply reading the documents takes significant
data such as the document title, author and date, to theoretical or time in and of itself—make sure you think about how, exactly, the
conceptual categories deriving from your research question, operat- specific data you are extracting will be used and how it goes towards
ing theory or analytical framework (Panel 2). Documents can also answering your research questions.
be imported into thematic coding software such as Atlas.ti or
NVivo, and data extracted that way. Alternatively, if the research Step 3. Analyse data
question focuses on process, documents can be used to compile a As in all types of qualitative research, data collection and ana-
timeline of events, to trace processes across time. Ask yourself, how lysis are iterative and characterized by emergent design, meaning
can I organize these data in the most coherent manner? What are my that developing findings continually inform whether and how to
priority categories? We have included two different examples of obtain and interpret data (Creswell, 2013). In practice, this
data extraction tools in the Supplementary Materials to this article means that during the data extraction phase, the researcher is al-
to spark ideas. ready analysing data and forming initial theories—as well as po-
Document analyses are first and foremost exercises in close read- tentially modifying document selection criteria. However, only
ing: documents should be read thoroughly, from start to finish, when data extraction is complete can one see the full picture.
including annexes, which may seem tedious but which sometimes For example, are there any documents that you would have
produce golden nuggets of information. Read for overall meaning as expected to find, but did not? Why do you think they might be
you extract specific data related to your research question. As you missing? Are there temporal trends (i.e. similarities, differences or
go along, you will begin to have ideas or build working theories evolutions that stand out when documents are ordered chrono-
about what you are learning and observing in the data. We suggest logically)? What else do you notice? We provide a list of over-
capturing these emerging theories in extended notes or ‘memos,’ as arching questions you should think about when viewing your
used in Grounded Theory methodology (Charmaz, 2006); these can body of document as a whole (Table 2).
be useful analytical units in themselves and can also provide a basis In addition to the meaning-making processes you are already
for later report and article writing. engaged in during the data extraction process, in most cases, it will
As you read more documents, you may find that your data ex- be useful to apply specific analysis methodologies to the overall cor-
traction tool needs to be modified to capture all the relevant infor- pus of your documents, such as policy analysis (Buse et al., 2005).
mation (or to avoid wasting time capturing irrelevant information). An array of analysis methodologies can be used, both quantitative
This may require you to go back and seek information in documents and qualitative, including case study methodology, thematic content
1430 Health Policy and Planning, 2020, Vol. 35, No. 10

analysis, discourse analysis, framework analysis and process tracing, Conclusion

which may require differing levels of familiarity and skills to apply
Document analysis is an essential component of health policy re-
(we highlight a few of these in the case studies below). Analysis can search—it is also relatively convenient and can be low cost. Using an
also be structured according to theoretical approaches. When it organized system of analysis enhances the document analysis’s pro-
comes to analysing policies, process tracing can be particularly use- cedural rigour, allows for a fuller understanding of policy process
ful to combine multiple sources of information, establish a chronicle and content and enhances the effectiveness of other methods such as
of events and reveal political and social processes, so as to create a interviews and non-participant observation. We propose the READ
narrative of the policy cycle (Yin, 1994; Shiffman et al., 2004). approach as a systematic method for interrogating documents and
Practically, you will also want to take a holistic view of the docu- extracting study-relevant data that is flexible enough to accommo-
ments’ ‘answers’ to the questions or analysis categories you applied date many types of research questions. We hope that this article
during the data extraction phase. Overall, what did the documents encourages discussion about how to make best use of data from
‘say’ about these thematic categories? What variation did you find documents when researching health policy questions.
within and between documents, and along which axes? Answers to

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

these questions are best recorded by developing notes or memos,
which again will come in handy as you write up your results. Supplementary Data
As with all qualitative research, you will want to consider your Supplementary data are available at Health Policy and Planning online.
own positionality towards the documents (and their sources and
authors); it may be helpful to keep a ‘reflexivity’ memo documenting
Acknowledgements
how your personal characteristics or pre-standing views might influ-
ence your analysis (Watt, 2007). The data extraction tool in the Supplementary Materials for the iCCM case
study (Panel 2) was conceived of by the research team for the multi-country
study ‘Policy Analysis of Community Case Management for Childhood and
Step 4. Distil your findings Newborn Illnesses’. The authors thank Sara Bennett and Daniela Rodriguez
You will know when you have completed your document review for granting permission to publish this tool. S.M. was supported by The
when one of the three things happens: (1) completeness (you feel sat- Olympia-Morata-Programme of Heidelberg University. The funders had no
role in the decision to publish, or preparation of the manuscript. The content
isfied you have obtained every document fitting your criteria—this
is the responsibility of the authors and does not necessarily represent the
is rare), (2) out of time (this means you should have used more spe-
views of any funder.
cific criteria), and (3) saturation (you fully or sufficiently understand
the phenomenon you are studying). In all cases, you should strive to Conflict of interest statement. None declared.
Ethical approval. No ethical approval was required for this study.
make the third situation the reason for ending your document re-
view, though this will not always mean you will have read and ana-
lysed every document fitting your criteria—just enough documents References
to feel confident you have found good answers to your research
Abdelmutti N, Hoffman-Goetz L. 2009. Risk messages about HPV, cervical
questions.
cancer, and the HPV vaccine Gardasil: a content analysis of Canadian and
Now it is time to refine your findings. During the extraction U.S. national newspaper articles. Women & Health 49: 422–40.
phase, you did the equivalent of walking along the beach, noticing Adebiyi BO, Mukumbang FC, Beytell A-M. 2019. To what extent is fetal alco-
the beautiful shells, driftwood and sea glass, and picking them up hol spectrum disorder considered in policy-related documents in South
along the way. During the analysis phase, you started sorting these Africa? A document review. Health Research Policy and Systems 17:
items into different buckets (your analysis categories) and building Atkinson PA, Coffey A. 1997. Analysing documentary realities. In: Silverman
increasingly detailed collections. Now you have returned home from D (ed). Qualitative Research: Theory, Method and Practice. London: SAGE.
Bardach E, Patashnik EM. 2015. Practical Guide for Policy Analysis: The
the beach, and it is time to clean your objects, rinse them of sand
Eightfold Path to More Effective Problem Solving. Los Angeles: SAGE.
and preserve only the best specimens for presentation. To do this, Bennett S, Dalglish SL, Juma PA, Rodrı́guez DC. 2015. Altogether now. . .
you can return to your memos, refine them, illustrate them with understanding the role of international organizations in iCCM policy trans-
graphics and quotes and fill in any incomplete areas. It can also be fer. Health Policy and Planning 30: ii26–35.
illuminating to look across different strands of work: e.g. how did Berner-Rodoreda A, Bärnighausen T, Kennedy C et al. 2018. From doxastic to
the content, style, authorship, or tone of arguments evolve over epistemic: a typology and critique of qualitative interview styles.
time? Can you illustrate which words, concepts or phrases were Qualitative Inquiry 26: 291–305. 1077800418810724.
Bowen GA. 2009. Document analysis as a qualitative research method.
used by authors or author groups?
Qualitative Research Journal 9: 27–40.
Results will often first be grouped by theoretical or analytic cat-
Bryman A. 1994. Analyzing Qualitative Data.
egory, or presented as a policy narrative, interweaving strands from Buse K, Mays N, Walt G. 2005. Making Health Policy. New York: Open
other methods you may have used (interviews, observation, etc.). It University Press.
can also be helpful to create conceptual charts and graphs, especially Charmaz K. 2006. Constructing Grounded Theory: A Practical Guide through
as this corresponds to your analytical framework (Panels 1 and 2). If Qualitative Analysis. London: SAGE.
you have been keeping a timeline of events, you can seek out any Claassen L, Smid T, Woudenberg F, Timmermans DRM. 2012. Media cover-
missing information from other sources. Finally, ask yourself how age on electromagnetic fields and health: content analysis of Dutch news-
paper articles and websites. Health, Risk & Society 14: 681–96.
the validity of your findings checks against what you have learned
Creswell JW. 2013. Qualitative Inquiry and Research Design. Thousand
using other methods. The final products of the distillation process
Oaks, CA: SAGE.
will vary by research study, but they will invariably allow you to Dalglish SL, Rodrı́guez DC, Harouna A, Surkan PJ. 2017. Knowledge and
state your findings relative to your research questions and to draw power in policy-making for child survival in Niger. Social Science &
policy-relevant conclusions. Medicine 177: 150–7.
Health Policy and Planning, 2020, Vol. 35, No. 10 1431

Dalglish SL, Surkan PJ, Diarra A, Harouna A, Bennett S. 2015. Power and Ministère de la Santé Publique. 2010. Enquête nationale sur la survie des
pro-poor policies: the case of iCCM in Niger. Health Policy and Planning enfants de 0 à 59 mois et la mortalité au Niger 2010. Niamey, Niger: MSP.
30: ii84–94. Mogalakwe M. 2009. The documentary research method—using documen-
Entman RM. 1993. Framing: toward clarification of a fractured paradigm. tary sources in social research. Eastern Africa Social Science Research
Journal of Communication 43: 51–8. Review 25: 43–58.
Fournier G, Djermakoye IA. 1975. Village health teams in Niger (Maradi Nelkin D. 1991. AIDS and the news media. The Milbank Quarterly 69:
Department). In: Newell KW (ed). Health by the People. Geneva: WHO. 293–307.
Gurung G, Derrett S, Gauld R. 2020. The role and functions of community Ouedraogo O, Doudou MH, Drabo KM et al. 2020. Policy overview of the
health councils in New Zealand’s health system: a document analysis. The multisectoral nutrition planning process: the progress, challenges, and les-
New Zealand Medical Journal 133: 70–82. sons learned from Burkina Faso. The International Journal of Health
Harvey L. 1990. Critical Social Research. London: Unwin Hyman. Planning and Management 35: 120–39.
Harvey SA. 2018. Observe before you leap: why observation provides critical Prior L. 2003. Using Documents in Social Research. London: SAGE.
insights for formative research and intervention design that you’ll never get Shiffman J, Stanton C, Salazar AP. 2004. The emergence of political priority
from focus groups, interviews, or KAP surveys. Global Health: Science and for safe motherhood in Honduras. Health Policy and Planning 19: 380–90.
Practice 6: 299–316. Smith KC, Washington C, Welding K et al. 2017. Cigarette stick as valuable

Downloaded from https://academic.oup.com/heapol/article/35/10/1424/5974853 by guest on 05 June 2022

Institut National de la Statistique. 2010. Rapport National sur les Progrès vers communicative real estate: a content analysis of cigarettes from 14
l’atteinte des Objectifs du Millénaire pour le Développement. Niamey, Niger: INS. low-income and middle-income countries. Tobacco Control 26: 604–7.
Kamarulzaman A. 2013. Fighting the HIV epidemic in the Islamic world. Strömbäck J, Dimitrova DV. 2011. Mediatization and media interventionism:
Lancet 381: 2058–60. a comparative analysis of Sweden and the United States. The International
Katchmarchi AB, Taliaferro AR, Kipfer HJ. 2018. A document analysis of Journal of Press/Politics 16: 30–49.
drowning prevention education resources in the United States. International UNICEF. 2010. Maternal, Newborn & Child Surival Profile. Niamey, Niger:
Journal of Injury Control and Safety Promotion 25: 78–84. UNICEF
Krippendorff K. 2004. Content Analysis: An Introduction to Its Methodology. Watt D. 2007. On becoming a qualitative researcher: the value of reflexivity.
SAGE. Qualitative Report 12: 82–101.
Marten R. 2019. How states exerted power to create the Millennium Weber M. 2015. Bureaucracy. In: Waters, TWaters, D (eds). Rationalism and
Development Goals and how this shaped the global health agenda: lessons Modern Society: New Translations on Politics, Bureaucracy, and Social
for the sustainable development goals and the future of global health. Stratification. London: Palgrave MacMillan.
Global Public Health 14: 584–99. Wesley JJ. 2010. Qualitative Document Analysis in Political Science.
Maxwell JA. 2005. Qualitative Research Design: An Interactive Approach, World Health Organization. 2006. Country Health System Fact Sheet 2006:
2nd edn. Thousand Oaks, CA: Sage Publications. Niger. Niamey, Niger: WHO.
Mayring P. 2004. Qualitative Content Analysis. In: Flick U, von Kardorff E, Yin R. 1994. Case Study Research: Design and Methods. Thousand Oaks, CA: Sage.
Steinke I (eds). A Companion to Qualitative Research. SAGE.