Presentation on theme: "Methodological Issues in Mixed Methods Data:"— Presentation transcript:
1 Methodological Issues in Mixed Methods Data: The Use of Qualitative and Quantitative Data in Health Services Research.Susan Zickmund, PhDDirector, Qualitative Research CoreCHERP, VA PittsburghDivision of General Internal MedicineUniversity of Pittsburgh
2 Goal for the Cyber Seminar Briefly describe the ABCs of qualitative research.Provide an introductory guide to mixed methods designs.Suggest practical ideas on how to best transform qualitative themes into numerical information and then to integrate that into a final analysis plan.
3 Organization of the Seminar Traditional qualitative: description, data types, methods, sample size, recruitment, coding.Mixed methods: traditional designs, an “Integrated Mixed Methods” approach (with methods, sample size, recruitment, coding), and conclude with examples of integrated mixed statistical models.
5 Description of Qualitative The focus is on the participants’ subjective viewpoints.Words/images are the primary data elements.The approach is inductive.Theory development may be a main outcome of the analysis.
6 Central Characteristics Having an iterative /open ended approach to the data.Observations or coding schemes emerge directly from the text.The researcher strives to avoid bias when interpreting the data.
7 Classic ComponentsTypes of data, qualitative methods, determining sample size, recruitment, and coding philosophies.
8 Data TypesInterviews/Focus Groups.Observational.
9 Interviews / Focus Groups Interviews: Allow in-depth discussion with one participant; effective for sensitive topics; interviewer controls the discussion.Focus Group: Allows participants to interact; group dynamics provide unique insight; moderator has less control over discussion with one participant.
10 Example: Hand washing study. ObservationalA complex situation where a researcher would need to observe what is occurring in order to best understand the situation.Example: Hand washing study.Participants may have reasons for dishonesty.The activity is open and observable.
11 Qualitative Methodologies Important to have one to guide data collection and analysis.Types include:Grounded Theory.Ethnography.
12 Grounded Theory Most prominent method in medicine. Uses constant comparisons between cases.Can change recruitment goals based on previous findings.The goal is an emerging theory.
13 EthnographyMethod of anthropologists.Involves field notes; goal to observe and understand a culture.Effective for an unique or unknown cultural dimension of medicine.Requires a research question best fitted for this method.
14 Sample SizeUses “thematic saturation” – idea that once no new themes arise, data collection is complete.Minimum sample size for saturation is aroundMaximum sample size is any size interfering with “case oriented thrust” of qualitative research (60-100).
15 Recruitment: Purposeful Sampling The goal of recruitment is to purposefully section special cases.It does not seek “generalizability.”Types of sampling used include:Extreme / maximum variation.Snowball sampling.
16 Coding Philosophies Single investigator. Research team approach. Independent coders.
18 Mixed Methods Terminology Multiple types of qualitative data or using experts with different academic backgrounds (“triangulation”).Newer: Integrating qualitative and quantitative data collection together.
19 The Qualitative-Quantitative Divide To some, qualitative is seen as incommensurate with empirical data.Thus, there is a need to conduct the qualitative study in a mixed methods design so as to best overcome this divide.
20 QUANT qual qual follow-up Mixed Methods DesignsTime and emphasis (in CAPS).qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL quantquant follow-up
21 QUANT qual qual follow-up Mixed Methods DesignsA smaller qualitative study designed to provide data for a larger quantitative one (often survey based).qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL quantquant follow-up
22 QUANT qual qual follow-up Mixed Methods DesignsA small quantitative study that is the set- up for the major qualitative study to follow.qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL quantquant follow-up
23 QUANT qual qual follow-up Mixed Methods DesignsA major quantitative study that uses qualitative data to gain insight into its findings.qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL quantquant follow-up
24 QUANT qual qual follow-up Mixed Methods DesignsA major qualitative study that uses a follow-up quantitative study at the end.qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL quantquant follow-up
25 Rarely Integrated at Analytic Level These studies are sequential.Participants infrequently complete all portions of the data.These cases do not lend themselves readily to analytic integration.
26 QUANT qual qual follow-up Simultaneous DesignWhere qualitative and quantitative methods reinforce simultaneously.qual QUANT qual preliminaryquant QUAL quant preliminaryQUANT qual qual follow-upQUAL QUANTsame time
27 Simultaneous: Gold Standard Quantitative (demographics, surveys, clinical) and qualitative data is collected from all participants.Analysis plan integrates the quantitative / qualitative data together.Few examples, but is the best method for fully interpreting data in an empirical study.
28 Integrating Mixed Methods (IMM): Overview Provide practical approach to:Research design.Analytic strategies that bestfacilitate integrating qualitativedata into empirical studies.
29 Data Type, Qualitative Methodology The data type used would be dictated by the research question and may not differ.One qualitative method that lends itself well is the quasi-statistical method by Crabtree and Miller*, a methodological approach developed for health research.Doing Qualitative Research, 1992.
30 Sample Size and Recruitment Sample size, rather than using “thematic saturation,” would be determined by sample size calculation.Recruitment, rather than using “purposeful sampling,” would be consistent with clinical research, using inclusion/exclusion criteria.Goal is generalizability.
31 Codebook Construction Use well-defined methods:Inclusion/exclusion criteria for codes.Sample ~20%-100% of cases for the codebook construction.Audit trail, time/date stamping.Goal is transparency of method.
32 Coding Philosophy Independent coders (two). Agreement model to adjudicate differences (need final master file).Inter-coder kappa statistic to measure reliability.Goal is reliability.
33 Interpreting Kappa Statistics 0.00 = poor= slight= fair= moderate= substantial= almost perfectRule of thumb: shoot for 0.70 and above
34 Data TransformationConvert thematic analysis into present / absent (0, 1).Convert thematic analysis into Likert scales.
35 Computer Data Management Software programs (Atlas.ti, Nudist) allow for computerized management of:Interview/focus group files.Codebooks.Codes.Enables a level of textual complexity not possible with notes alone.
37 Computerized Data Output Atlas includes the ability to output data to tables and spreadsheets (Excel, SPSS).Atlas uses 0/1 for presence and absence of codes (Likert scales require adaptation).
38 Computerized Data Output Useful for master files for integration with other data, and facilitates intercoder reliability files.MSTVIEWBETTERWORSENO CHANGESEC1C4362C4703C1074C695C16C477
39 IMM Summary *Sample size calculation over “saturation.” *Simultaneous mixed methods, where the data is collected from all participants.*Sample size calculation over “saturation.”*Recruitment: generalizability over “purposive sampling.”Transparency of codebook/coding methods.Intercoder reliability kappa statistics.Computerized management spreadsheets.
40 Examples of Using Qualitative Data in Statistical Models Qualitative data as a predictor variable.Qualitative data as an outcome variable.
41 The impact of chronic disease on cancer patients' self conception Patient Narrative StudyThe impact of chronic disease on cancer patients' self conceptionIMM: Simultaneous mixed methods design: 1 hour semi-structured interview, survey data, demographics, clinical data.
42 Qualitative Interview “View of Self” Code:“As you go through this experience, have you begun to think about yourself differently?”Prompts used to guide beyond yes/no.
43 Additional Quantitative Data Mortality data (current).Charleston Comorbidity data.Cancer Staging (time of interview).Demographics (self report).Sickness Impact Profile (sub-scales).Hospital Anxiety Depression Scale (Anxiety/Depression scores).
44 Distribution of Answers Data available on 825 participants for“View of Self” code:Better View %Unchanged View %Worse View %
45 Example of Better View of Self: “I'm back to realizing that I do have an internal strength; that it will take me wherever I need to go in this journey. And it will be a good journey, whatever the end outcome is.”
46 Example of Worse View of Self: “I am not the person that I was (cries). Just to grasp the concept that at a young age you're disabled, just like overnight, is very hard to swallow. That's a very hard thing to tell someone: `Too bad, your life is ruined, you just better learn to go on.’ And at 37, you're thinking: `Oh my gosh, I just had a baby.’ ”
47 Univariate Results for Predictors of Mortality “View of Self” <0.001Age <0.001Employed> High School educationCancer staging <0.001Sickness Impact ProfileAmbulationAppetitePhysical Sub-scales
48 Final Multivariable Model for Predictors of Mortality Variables p-valueView of selfAge <0.001Cancer Stage <0.001SIP-AppetiteSIP-Ambulation
49 Racial Disparities QI Project Racial Disparities in Satisfaction with VA CareZickmund SL, Burkitt KH, Rodriguez KL, Switzer GE, Stone RA, Shea JA, Gao S, Bayliss N, Meiksin R, McClenney LM, Powell CT, Newsome ES, Allen R, Fine MJ.Center for Health Equity Research and Promotion, VA Pittsburgh Healthcare System and Philadelphia VA Medical Center; VA Center for Minority Veterans (CMV); VHA Office of the Assistant Deputy Undersecretary for Health
50 BackgroundThe 2008 VHA Hospital Report Card revealed racial disparities in veteran satisfaction with VA health care.CHERP and the Center for Minority Veterans were commissioned to identify reasons for the disparity between African Americans and whites.
51 ObjectivesTo determine whether racial differences in satisfaction existed in overall, outpatient, and inpatient VA care.To describe racial differences in satisfaction in eight domains of health care quality.
52 DesignMulti-site QI of 30 white/30 African American veterans (20 per site).Telephone interviews with Likert scale and open-ended questions.Demographic data collected.
53 Qualitative MethodsInterviews were coded by 2 coders using an iteratively developed codebook.Intercoder reliability statistic was Kappa=0.99.Coded themes were then aggregated within the 8 health care domains, with a distinction between “satisfied” and “dissatisfied.”
54 8 Health Care Quality Domains Trust in providerPain managementFeelings of respectAccess to medical careCommunication with providersCoordination of careInvolvement of family and friends.Role of race
55 Domain: Access to Care“One of the things [that’s] a concern for me individually right now is that I’m trying to get a primary care doctor now, and that’s like, well…I haven’t had one, and I’ve been attending the VA off and on for six, seven years.”
56 Domain: Role of RaceVeteran: “A lot of times, they, especially people of color and black, Hispanics, Latino, et cetera like that, they [the providers] have a tendency to act like we’re lying, or you want to get high, or you’re trying—you know, it’s almost—you’ve got to either act out or cry or some kind of way to validate…”Q: “That you’re really in pain.”Veteran: “Yes, yes.”
57 Statistical Analysis of Qualitative Data 1. Using Chi Square statistics on codes.
58 Total(N=60)African American (N=30)White(N=30)p-valuen%Domain 1: Trust in Medical ProvidersDissatisfiedPrimary care provider201240.0826.70.40Provider(s) not giving enough information181136.7723.3
59 Statistical Analysis of Qualitative Data Using statistical modeling.Item response theory approach (the Rasch model).Fit random intercept logistic models were used to assess the differences between African American and white veterans accounting for domain and dissatisfaction/satisfaction themes.
60 Themes of Satisfaction and Dissatisfaction by Domain Odds Ratiofor African Americans(95% CI )P-valueTrust in Medical ProvidersDissatisfied1.40 (0.63, 3.09)0.41Satisfied0.36 (0.18, 0.73)0.005
61 ConclusionMixed methods allow the richness of qualitative themes to be used along with quantitative data.Sequential designs facilitate combining qualitative and quantitative work, but do so in a segmented way.IMM approach enables the integration of the qualitative data at the level of the statistical analysis.
62 Director, Qualitative Research Core Questions?Susan Zickmund, PhDDirector, Qualitative Research CoreCHERP, VA Pittsburgh