Using mixed methods approach in a health research setting Dr Caroline Bulsara, School of Primary, Aboriginal and Rural Health Care, University of Western Australia
Reasons why people prefer one methodology over another Quantitative Clear cut All methodology decided on before commencing Statistical results convincingly prove / disprove the hypothesis Qualitative It seems easier eg focus groups May be more relevant to field of employment Flexibility with process Lack of understanding of clinical / scientific research
Use of mixed methods approach Variation in data collection leads to greater validity. Answers the question from a number of perspectives. Ensures that there are no gaps to the information / data collected. Ensures that pre existing assumptions from the researcher are less likely. When one methodology does not provide all the information required.
Types of research using mixed methods Services orientated. Evaluative and needs analysis research. Translatable into new or changes in policy. Provides a broad spectrum approach to an issue. Uses qualitative exploratory data to formulate a quantitative study. Uses quantitative data to explore issues more qualitatively.
Background mixed methods Approach and main focus on either qualitative or quantitative will depend on popn. Instrument design model. Triangulation design model. Data transformation design model. Explanatory model.
THE INSTRUMENT DESIGN MODEL
Instrument design model Priority to quantitative data. Two phases begins with qualitative and then moves to quantitative design and testing. Integration in the data analysis stage. Uses the qualitative information to develop an instrument for data collection. Eg Scale development or Focus groups to create a questionnaire.
Exploring a topic to formulate a quantitative study: Example to develop a scale to measure patient empowerment. Wanted to quantify empowerment. The concept is not directly observeable. Need to operationalise the concept. Identifying the markers of empowerment. How was this done?
ATTITUDE STATEMENTS Serve as markers for empowerment. What the patient will believe / act out is either endorsed or denied for each statement. Needed the information to develop the statements.
STUDY DESIGN Extensive literature review. Motivation, patient information seeking, relationship with health professionals. In depth interviewing of Shared Care patients. Saturation point reached. Analysis and themes distilled. Development of statements.
WHO TO INTERVIEW? Patients. Their partner / family member. Prognosis less important than degree of empowerment. Diverse ages.
Information relating to illness Involvement in decision making process (DMP) Family support Support of friends Relationship with GP Patient perception of GP ability to manage illness Patient perception of health professionals willingness to include them in DMP EMERGING THEMES Complementary therapies Spiritual beliefs Acceptance and adaptability to illness Patient perceived usefulness to friends Patient perceived usefulness to family Paid employment Resources
DATA ANALYSIS Data analysed and themes distilled. Key themes identified and written up. Statements formulated. Statements reviewed by various experts and cancer patients. Questionnaire distributed.
MARKERS FOR EMPOWERMENT Resources Information relating to illness Capability of using resources to handle illness Sufficient resources to handle illness Sufficient information Relevance of information Involvement in decision making process Desire for involvement in decision making process Capability to be involved in decision making process Family support Availability of a supportive family Patient need for the support of family Support of friends Availability of supportive friends Patient need for the support of friends
TRIANGULATION DESIGN MODEL
Triangulation design model Frequently used in primary health care research. Gathers data at the same time. Integrate all data in order to clarify or better understand the problem. Quantitative and qualitative are given equal priority. Reported on in separate sections of a report. Discussion brings salient points together in the report.
Triangulation design model: example What are the issues in accessing services for those affected by traumatic injury? What information do we require? What evidence do we require to show that there are problems?
What groups /representatives? What data do we require? - Persons with Catastrophic injury (CI). - Carers (family) of persons with CI. - Service providers / orgs for those affected by CI. - Health professionals. What parameters? - Types of injury. - Level of care required. - Compensation? - Age groups. - Paid care? - Length of hospital stay. - Types of aids and appliances. - Impact on family.
Methods of data collection - What is appropriate for the following groups? Survey Database Focus groups Interviews telephone / personal Persons with Catastrophic injury (CI). Carers (family) of persons with CI. Service providers / orgs for those affected by CI. Health professionals.
Background work Terms of reference to be addressed from funder. Consult community organisations. Select steering panel. Read literature and find one contact person within organisations. Discuss appropriate methods for accessing target groups.
Using multiple methods What questions are appropriate for use in a survey? What is better asked in a confidential or one on one setting? What format will best utilise both researcher and participant s time?
Outcomes for traumatic injury study Several stakeholder groups discussion highlighted several key issues common across groups. The database findings provided background for the more in-depth information from the in depth interviews.
DATA TRANSFORMATION DESIGN MODEL
Data transformation design model Correlational (observational) designs prevalence studies, prospective studies. Favours quantitative data collection and analysis. Researcher analyses qualitative data and codes and numerically counts themes. Concurrent data collection. Lends well to evaluative (survey) type research.
Why evaluate? Compares what you observe with o criterion or standard of acceptability, or o standard of indicators of good performance Uses qualitative and quantitative methods to determine if program o developed and implemented as planned o met its goals and objectives
Evaluative research techniques Data collected by: Postal or mail surveys E-mail surveys Internet surveys Telephone surveys Door-to-door surveys Central location (e.g. Shopping centre) surveys Observational techniques Written material and resources Recruitment and conduct of group discussions, in depth interviews, focus groups.
Data transformation design model example An evaluation of an asthma awareness program asks new mothers by telephone interview what changes they have made to protect their baby from exposure to smoke. - Telephone interview and survey with open ended responses.
SHORTER TYPE OF RESPONSE
PRESENTATION OF SHORTER RESPONSE
SPSS TEXTSMART Import files as text documents (.txt) Prepare your data Create categories - either automatic or manually done by you Refine categories Visualise data - category plots and brushing Refine categories again Export results if needed
Respondent or interview number Response or question number Text to be included is hyphenated
WORD FREQUENCY CHART Number of times word or term was mentioned
Example content analysis presented qualitatively Women s beliefs about the causes of breast cancer (eg from Sue Wilkinson in Qualitative Psychology Ed Jonathon Smith).\ Category and then quote: 5. Stress, strain and worry. Not discussed. 6. Caused by childbearing I mean I don't know whether the age of which you have children makes a difference as well because my [pause] 8 year old relatively late, I was an old mum. They say if you ve had one, you re more likely to get it than if you ve had a big family. 7. Secondary to trauma or surgery
Content analysis quantitatively presented ATTRIBUTION OF ILL HEALTH Stress, strain and worry. 0 times. Caused by childbearing. 22 times. Secondary to trauma and surgery. 9 times. Allows the researcher to see how generalisable some themes are.
EXPLANATORY MODEL
Explanatory Model Initial quantitative phase. Statistical results. Second phase uses qualitative methods to explain or elucidate results.
Example 2 Mixed methods research Issues around senior s medication safety in WA. Statistical results showing non compliance amongst seniors using PBS data and hospital ED presentations. What research methods are relevant? How do we go about investigating this?
Background and initial work Don t really know what the issues are apart from other studies (literature). Need as broad a spectrum of perspectives as possible. A seniors forum advertised through community groups and newspapers. Forum and break out groups. Main issues identified. Steering panel, focus groups invitations, forum results written up. Data linkage project around PBS and Medicare items.
Broad Issues Raised Role of GP and pharmacist. Side effects and interactions. Generic brands (attitudes). Continuity of care. Chronic illness self management. Packaging. Over the counter meds.
Appropriate methodology What type of questions do you have? Surveys? Interviews? Focus groups? What is most appropriate for your target group(s)? What will initiate the greatest number of responses? Time and funding.
Method Decided on a forum and focus groups. Forum defined main issues for this group in ensuring medications safety. Separate focus group for each topic. Set of questions to be addressed per focus group.
Outcomes Findings revealed reason for high levels of non compliance amongst seniors. Also revealed issues that researchers were not previously aware of. Suggested ways that the community felt would assist in maintaining better compliance amongst seniors. Recommendations being drafted currently.
Conclusions Can be labour intensive. Involves multiple stages of data collection. Provides greater breadth of perspectives around a certain issue. Combining approaches helps overcome deficiencies in one method only. Effectively allows populations with limited language skills or trust issues to participate. Can help define more nebulous concepts. Prevents researcher assumptions about a particular population. Lends itself well to outcomes driven research (eg needs assessments or evaluations).