HLTH 8202E Week 9 Survey Methods Section Example

Methods: A Cross-Sectional Survey of Paid Home Care Aides on Heat During Client Visits, Reported in the Order a Replication Would Need, From Population and Two-List Frame to Text-to-Web Contact and AAPOR Response Rates

Student Name

Walden University

HLTH 8202E: Survey Research Methods

Instructor Name

Month Day, Year

Methods: A Cross-Sectional Survey of Paid Home Care Aides on Heat During Client Visits, Reported in the Order a Replication Would Need, From Population and Two-List Frame to Text-to-Web Contact and AAPOR Response Rates

Design

The study is a cross-sectional survey of paid home care aides in a composite state, conducted in the autumn after the summer it asks about. It measures three constructs defined before item writing: exposure to hot conditions during client visits, control over those conditions and physical strain during or after hot visits. The survey was developed over a term of documented steps, and each decision below cites the document in which it was argued.

Population

The target population is every paid aide who gave personal or health-related care inside a client's home in the state during June, July or August of the reference year. Membership depends on the job done rather than on who paid for it, and includes aides employed through licensed agencies, through the state's consumer-directed program, privately by families and through more than one of these routes.

Frame and Coverage

The frame combines two lists: rosters from licensed home care agencies that agreed to share them under a data agreement, and the payroll file of the consumer-directed program's fiscal agent. Agencies were asked to include anyone employed during the reference summer, rather than just the names still listed in the fall, to reduce undercoverage of aides who left. The frame does not cover aides at agencies that declined to share rosters, aides hired privately by families or aides missing from both lists. These groups are described in the sampling plan, and their likely direction of bias is discussed in the limitations.

Sampling

The sample was stratified by agency size, in three groups, and by employment route, in two groups, to guarantee places in the sample for small agencies and for aides paid through the consumer-directed program. Within each stratum, aides were selected at random from the combined list. Where permitted, the two lists were linked on personal identifiers so that aides already known to appear twice could be marked before any draw. Because some aides appear on both lists without a detectable match, the questionnaire asks each respondent which routes they worked through, and aides reporting both receive a weight reflecting their two chances of selection, following the logic of multiple-frame estimation (Lohr & Rao, 2006).

Instrument Development

The instrument was built in four stages. A construct memo defined exposure, control and strain, with a fixed recall period and the aide's vantage point. An item bank drafted more items than the final survey would use, each tied to one construct and one analysis. An expert review read the draft against the four stages of the survey response process, and a scale review revised every option set, adding a no-visits category, labeling frequency points and replacing agree-disagree formats with item-specific questions. The final instrument, in the appendix, contains 22 items and takes about eight minutes to complete on a phone.

Pretesting

Two rounds of cognitive interviews, 15 and then 10, combined think-aloud with scripted probes for four suspect items (Willis, 2005). Interviews were conducted in English and in two other languages with aides recruited by language, employment route and experience. The first round led to three revisions: "too hot to work in comfortably" was anchored to the aide's own comfort; "break" was replaced with a description of pausing work; and a separate path was added for live-in aides, who could not count visits. The second round found no item misread by three or more respondents. A field pretest with 40 aides by text and web checked completion time and item missing data.

Translation

The questionnaire was translated into the two non-English languages that aides in the state most often prefer by forward translation, review by a bilingual panel that included aides, and adjudication of disagreements. Translated versions were pretested in their own languages. Changes made in one language were carried into all versions to keep the source and translations aligned.

Mode and Contact

The survey was fielded by text message with a web link, with telephone follow-up for nonrespondents, because aides are more reliably reached by phone than by mail or work email. The contact sequence followed Dillman's tailored design method, which uses multiple contacts, varied in timing and approach, to build trust and reduce the burden of responding (Dillman et al., 2014). The sequence was a pre-notice from the aide's agency or program, an invitation text with the link, two reminder texts several days apart and a telephone call to remaining nonrespondents. Respondents received a $20 electronic gift card on completion.

Response Rates

Response rates are reported using the American Association for Public Opinion Research's standard definitions (American Association for Public Opinion Research, 2023). The main rate reported is Response Rate 3, which estimates the share of cases of unknown eligibility that are eligible, because some numbers on the lists belonged to people who were not aides last summer. Response Rate 1, which treats all unknown cases as eligible, is also reported. Naming the definition allows readers to compare the rate with other surveys, since a single unlabeled percentage can be computed in many ways.

Weighting

Base weights reflect each aide's probability of selection, adjusted for multiple-frame membership. Nonresponse adjustment uses agency size, employment route and years of experience, which are available on the frame for respondents and nonrespondents alike.

Analysis

The main analysis estimates the share of aides in each exposure category and the frequency of strain symptoms, with confidence intervals reflecting the stratified design and weights. A second analysis tests whether control over cooling is associated with lower strain at the same level of exposure, using a weighted regression of symptom frequency on exposure, control and background variables. Open-text answers to "other" categories are coded by two analysts, who resolve disagreements by discussion and report how often they first disagreed. Item missing data are reported for every item; where missingness on a key item exceeds 5%, results are shown with and without imputation so readers can see whether the choice matters.

Limitations

The frame misses privately hired aides and those at nonparticipating agencies, who may face more exposure and less control; estimates may therefore understate exposure. All measures are self-reported and subject to recall error, although the bounded summer window and repeated recall period in each stem aim to reduce it. The survey is cross-sectional, so the association between control and strain does not establish cause.

References

American Association for Public Opinion Research. (2023). Standard definitions: Final dispositions of case codes and outcome rates for surveys (10th ed.). https://aapor.org/standards-and-ethics/standard-definitions/

Dillman, D. A., Smyth, J. D., & Christian, L. M. (2014). Internet, phone, mail, and mixed-mode surveys: The tailored design method (4th ed.). Wiley.

Lohr, S., & Rao, J. N. K. (2006). Estimation in multiple-frame surveys. Journal of the American Statistical Association, 101(475), 1019-1030. https://doi.org/10.1198/016214506000000195

Willis, G. B. (2005). Cognitive interviewing: A tool for improving questionnaire design. Sage.