GB891 Unit 10 Implementation Chapter Draft Example

What Actually Happened in Twelve Weeks: A Unit 10 Implementation Chapter Draft for a Bakery Safety Study

[Student Name]

Purdue University Global

GB891: Doctoral Project II: Project Design and Implementation

Unit 10 Implementation Chapter Draft

[Instructor Name]

[Date]

Corliss Baking Company, its plants, participants and every figure are composite and illustrative; the methods sources cited are real.

Section 3: Implementation

This section describes how the study was carried out, compared with the protocol approved before it began. It reports participation, fidelity of the intervention, data collection completeness, observer agreement and departures from the protocol. Outcomes are reported in Section 4.

Participation

All 24 study crews, twelve at the intervention plants of Ashford and Brierly and twelve at the comparison plants of Calder and Dunmore, remained in the study for its full twelve weeks. At baseline, the crews included 412 associates, of whom 171, or 42 percent, were placed by the two staffing agencies. The baseline climate survey was returned by 264 associates, a response rate of 64 percent. Twenty-two of the 24 crews reached the eight-response minimum needed to compute a crew score; the two that did not were small night crews at Calder, which are reported but excluded from crew-level comparisons.

Turnover was substantial, as expected. Over the twelve weeks, 156 associates joined study crews and 149 left, so that across the intervention crews about 38 percent of members at the end of the study had not been on the crew at baseline. The follow-up survey was returned by 231 associates, 58 percent of those on the crews in week twelve, and 21 crews met the floor.

All twelve intervention supervisors and ten of the twelve comparison supervisors completed the week-thirteen interview; one comparison supervisor declined and one had left the company.

Fidelity of the Intervention

Fidelity is reported against the thresholds set before the study, following the elements Carroll et al. (2007) describe.

Delivery. Of 144 planned supervisor-weeks, 137 had a sheet sent. The seven without a sheet were the weeks after one Brierly night supervisor resigned in week six. Of the 137 sheets sent, 113 were on time and 24 were late, all in weeks four and ten. No sheet was missed, and all randomly checked sheets matched the pulse data.

Receipt. Of the 137 sheets sent, 126 were opened by the end of the Monday shift, 9 were opened later in the week and 2 were never opened. The two unopened sheets belonged to one Ashford supervisor in weeks two and three; after the research assistant's contact, the supervisor asked for a printed copy, and every later sheet was received.

Contamination. One contamination event occurred: a plant-wide safety stand-down at Dunmore in week nine, held by a new manager after a near miss.

Data Collection Completeness

Pulse draws were completed in 286 of 288 planned crew-weeks, yielding 1,089 of a targeted 1,152 answers, 95 percent. The two missing crew-weeks were the small Calder night crews in week seven, when a snowstorm closed the roads on both draw nights and the research assistant could not reach the plant. The shortfall in answers came mainly from night crews with fewer than four members present.

Observation sessions were completed as scheduled at three plants. At Ashford, one week of observations, week seven, was not conducted because of observer illness.

Observer Agreement

Joint observation sessions were conducted every fourth session throughout. Across all stations, the kappa statistic Cohen (1960) introduced averaged 0.78, a level Landis and Koch (1977) would call substantial agreement. The revised jam-clearing definition, introduced after the pilot, produced kappa of 0.71 at that station, lower than the others but above the moderate threshold of 0.60 that the log used to flag problems.

The feedback reached its supervisors in 137 of 144 planned weeks, which is what makes any result in Section 4 interpretable.

Departures From the Protocol

Five departures occurred, documented in full in Appendix F. Each is summarized here with the measure it affects. The Brierly supervisor resignation removed feedback from one crew for seven weeks and affects that crew's follow-up climate score. The Dunmore stand-down narrowed the contrast for near-miss reporting at that plant. The missed observation week reduced Ashford's compliance data by one week. The two late cycles affected 24 supervisor-weeks of delivery. And the merged crew affected one shift. As the deviation report concluded, each departure could only shrink an apparent effect, never inflate it.

How Supervisors Responded

The week-thirteen interviews, analyzed fully in Section 4, gave an early picture of how the intervention was received. Most intervention supervisors said they read the sheet on Monday morning before the first shift meeting and that the crew averages surprised them in the first weeks, because they had believed they talked about safety more than their crews reported. Several described changing small habits, such as mentioning a hazard when assigning a station or asking a new associate about the slicer guard during the first hour. Three said the sheet felt like being graded at first and that the feeling eased once they saw it never reached their plant managers. Two said they found the numbers hard to act on when their crew changed from week to week, which is the condition the study was designed to examine. One supervisor at Ashford said she had started keeping a short list of who was new each week so that she could make a point of talking with them first.

Newcomers and Agency Associates in the Pulse Data

Because the pulse draws were random, newcomers and agency associates were drawn roughly in proportion to their presence. Of the 1,089 pulse answers, 41 percent came from agency associates and 29 percent from newcomers within ninety days of joining a crew. Both shares are close to their shares of crew membership, which means the pulse data represent the high-turnover crews the study concerns rather than only their stable cores.

What This Means for the Results

The intervention was largely delivered as designed: 95 percent of planned supervisor-weeks received a sheet, and 92 percent of sheets sent were opened on the day they arrived. The study's central condition, high turnover, was present: more than a third of intervention crew members at the end were new. The concern Shadish et al. (2002) raise about history was partly addressed by the timing, but the Dunmore stand-down shows that events within the study period still occurred. Section 4 reports outcomes with these conditions in view, beginning with the climate survey and then the weekly pulse and near-miss series, including analyses restricted to on-time delivery weeks and comparisons with and without Dunmore.

References

Carroll, C., Patterson, M., Wood, S., Booth, A., Rick, J., & Balain, S. (2007). A conceptual framework for implementation fidelity. Implementation Science, 2, Article 40. https://doi.org/10.1186/1748-5908-2-40

Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46. https://doi.org/10.1177/001316446002000104

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159-174. https://doi.org/10.2307/2529310

Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Houghton Mifflin.