DDHA 8303 Week 11 Methodology Section Example

Methodology for a Mixed Methods Study of Coding Accuracy After a Move to Remote Work: Design, Coders, Measures, Procedures, Analysis and Safeguards in One Account

Student Name

Walden University

DDHA 8303: Research Methods for Healthcare Administrators

Instructor Name

Month Day, Year

Methodology for a Mixed Methods Study of Coding Accuracy After a Move to Remote Work: Design, Coders, Measures, Procedures, Analysis and Safeguards in One Account

Problem and Research Questions

Halvorsen Health, a composite five-hospital system, shifted 41 of its inpatient coders to home-based work during 2024, some in January and the rest in September, while the others continued onsite or hybrid. Leaders have since attributed a rise in payer DRG downgrades to the change, but no one has examined whether coding accuracy, as measured by the system's own audits, actually shifted. This study will address three questions. The first concerns whether audit agreement on the assigned DRG changed for coders after they moved home, relative to the change among coders who had not moved in the same months. The second concerns whether remote and onsite coders differ in work engagement and whether engagement tracks any change in agreement. The third concerns how coders describe the way they resolve uncertain coding decisions, and whether remote and onsite coders describe that work differently.

Research Design

The study will use an explanatory sequential mixed methods design. The quantitative strand, which answers the first two questions, will be completed before the qualitative strand begins, so that interviews can probe whatever pattern the audit analysis finds. The first question will be addressed with a difference-in-differences design with staggered timing, in which coders who moved in January 2024, coders who moved in September 2024 and coders who never moved are compared over 36 months. This design was chosen over a single cross-sectional comparison, which could not separate the effect of moving from differences in who was allowed to move, and over a single-group time series, which could not separate the move from changes affecting all coders, such as the annual ICD-10 update and the computer-assisted coding software adopted in 2024. Its central assumption, that agreement among movers and non-movers would have followed the same course had nobody gone home, will be examined through the months before each move (Wing et al., 2018).

Setting and Population

The setting is the inpatient coding department of Halvorsen Health, which serves five hospitals and audits a random sample of each coder's inpatient accounts every month. The population is every person who held the Inpatient Coder I or II job code during the three calendar years 2023, 2024 and 2025. Matching human resources records, the coding workflow system and the audit database identifies 78 such people, 64 still employed at the end of 2025.

Sampling

The quantitative strand will include all 78 coders who meet the inclusion rules, namely a run of six or more months, each with at least ten audited accounts, and a work arrangement on file for each of those months. Because the design compares each coder with themselves over time, a census is preferred to a sample. The engagement survey will be offered to all current coders. The qualitative strand will select up to sixteen current coders purposefully for maximum variation in move wave, tenure and hospital of origin, as Palinkas et al. (2015) describe for studies seeking a range of experiences. Nine agency coders who coded inpatient accounts during the period are excluded because their work arrangements were not recorded, and the fourteen coders who left cannot be interviewed. The study will seek de-identified exit interview summaries for those fourteen and will compare survey respondents and interviewees with the full population on tenure, wave and 2023 agreement, since nonresponse matters most when it is related to what is being measured (Groves & Peytcheva, 2008).

Instrumentation

Coding accuracy will be measured by the internal audit process already in place: an auditor independently assigns a DRG to each sampled account, and agreement is recorded as a yes or no. The study will not alter this process. Work engagement will be measured with the nine-item Utrecht Work Engagement Scale (Schaufeli et al., 2006), scored as a total from 0 to 6 because of the high correlations among its three subscales and mixed evidence about their separate structure. Coders will be told to think of the last three months when answering, which differs from the published administration and will be reported as such. Interviews will follow a semi-structured guide of eight questions, piloted with two coding supervisors who are not participants, asking coders to walk through recent records they found difficult and describe how they reached a decision.

Data Collection Procedures

Data collection will proceed in four stages. First, the research office will receive a de-identified extract of monthly audit results and work arrangement histories under a data use agreement, with names replaced by study codes. Second, the research office will send the survey invitation to all current coders, open the survey for three weeks and send one reminder. Third, after the quantitative analysis is complete, the research office will invite current coders to interview, screen respondents against the variation targets using frame data and schedule 45-minute interviews during paid time excluded from productivity tracking. Fourth, interviews will be recorded, transcribed, checked against the recording and the recording then destroyed.

Data Analysis

For the first question, monthly agreement proportions weighted by the number of audited accounts will be analyzed with a group-time estimator suited to staggered timing (Callaway & Sant'Anna, 2021), producing wave-specific and overall estimates and a month-by-month profile around each move. Months before each move will be examined for pre-existing differences in trend. Sensitivity analyses will exclude the month before each move, exclude months with fewer than ten audited accounts and restrict the sample to coders present for all 36 months. A logistic mixed model at the account level will serve as a check on the main estimate. For the second question, engagement scores will be compared by arrangement in a regression adjusting for tenure, with internal consistency reported by group; the association between engagement and change in agreement will be reported as exploratory. For the third question, the framework method (Gale et al., 2013) will be applied to the transcripts, combining codes drawn from published sources of coding error with codes developed from the data, and charting results in a matrix that separates remote from onsite coders. The strands will be integrated in a joint table placing each wave's estimate beside its interview themes.

Ethical Considerations

The review board will be asked to waive individual consent for the de-identified audit records and to approve a consent process for the survey and interviews in which invitations come from the research office rather than coding managers. No individual-level data will be shared with Halvorsen managers, and a written agreement bars use of study data for performance management. Results will be reported by wave and arrangement, never by hospital, and no group smaller than five will be shown. Coders may withdraw interviews within thirty days, after which quotations can be removed but the interview's contribution to the analysis cannot.

Limitations

Some limits are built into the design. The study cannot rule out that the moved and comparison groups would have diverged for reasons unrelated to remote work, though the pre-move months allow this to be examined. It cannot fully exclude spillover, since home-based coders may have turned to office colleagues for advice less often, which would touch both groups. Its findings apply to one organization's coders and one audit process. Other limits come from access. Agency coders are excluded, departed coders can contribute only records and possibly exit summaries, and the survey sample of about 64 is too small to test whether the engagement items work the same way for remote and onsite coders. These limits will be stated wherever the findings are reported.

References

Callaway, B., & Sant'Anna, P. H. C. (2021). Difference-in-differences with multiple time periods. Journal of Econometrics, 225(2), 200-230. https://doi.org/10.1016/j.jeconom.2020.12.001

Gale, N. K., Heath, G., Cameron, E., Rashid, S., & Redwood, S. (2013). Using the framework method for the analysis of qualitative data in multi-disciplinary health research. BMC Medical Research Methodology, 13, Article 117. https://doi.org/10.1186/1471-2288-13-117

Groves, R. M., & Peytcheva, E. (2008). The impact of nonresponse rates on nonresponse bias: A meta-analysis. Public Opinion Quarterly, 72(2), 167-189. https://doi.org/10.1093/poq/nfn011

Palinkas, L. A., Horwitz, S. M., Green, C. A., Wisdom, J. P., Duan, N., & Hoagwood, K. (2015). Purposeful sampling for qualitative data collection and analysis in mixed method implementation research. Administration and Policy in Mental Health and Mental Health Services Research, 42(5), 533-544. https://doi.org/10.1007/s10488-013-0528-y

Schaufeli, W. B., Bakker, A. B., & Salanova, M. (2006). The measurement of work engagement with a short questionnaire: A cross-national study. Educational and Psychological Measurement, 66(4), 701-716. https://doi.org/10.1177/0013164405282471

Wing, C., Simon, K., & Bello-Gomez, R. A. (2018). Designing difference in difference studies: Best practices for public health policy research. Annual Review of Public Health, 39, 453-469. https://doi.org/10.1146/annurev-publhealth-040617-013507