HD520 Unit 8 Data Analysis Section Example

Twenty-One Points, Read as an Interval: A Unit 8 Data Analysis Section for a Refugee Health Orientation Assessment

[Student Name]

Purdue University Global

HD520: Health Education Program Assessment

Unit 8 Data Analysis Section

[Instructor Name]

[Date]

The agency, the cohorts and every result are composite; the methods sources on clustering, intervals and nonequivalent comparison groups cited are real.

Sample

Of [268] adults on the roster, 241 agreed to take part and completed the baseline task: 131 early starters and 110 adults waiting for a seat. By day 42, 55 adults (23 percent) had been lost, leaving 101 early starters and 85 waiting adults in the analysis. Losses were similar in the two groups, 30 and 25 adults.

Baseline Comparison

Before reading any outcome, the analysis checks whether the cohorts started out alike, because a difference at day 42 means little if the groups were unequal on arrival.

Table 1 compares the analyzed cohorts. Mean age was 34.2 years among early starters and 33.5 among waiting adults. Median years of schooling were 8 and 7. The share who had visited any clinic since arrival was 18 and 21 percent. Baseline task completion was 31 percent (31 of 101) and 29 percent (25 of 85).

One imbalance stands out. Swahili speakers made up 9 percent of early starters (9 adults) but 22 percent of waiting adults (19 adults). Most Swahili-speaking families arrived during a spring surge when every class was full, so they waited. Dari speakers were 41 percent of early starters and 33 percent of waiting adults; Arabic speakers 35 and 28 percent; adults using telephone interpreters 16 percent in each group.

Main Result

Table 2 reports task completion at baseline and day 42. Among early starters, completion rose from 31 to 64 percent (65 of 101), a gain of 33 points. Among waiting adults it rose from 29 to 41 percent (35 of 85), a gain of 12 points. Waiting adults also improved, which matches what the Unit 2 design expected: newly arrived adults learn from caseworkers, relatives and their own first contacts with clinics.

The comparison of these gains, a difference in differences, was 22 points before adjustment. Adjusted for language group and for clustering of early starters within their twelve classes, it was 21 points, and its 95 percent interval ran from 6 to 36 points. Both groups were tested by the same assessors using the same script, so any practice effect from the first attempt should apply to each. Figure 1 plots the two groups' completion at both time points.

Clustering matters because adults taught together share an educator, a room and a set of classmates, so their results are more alike than those of adults taught separately. Ignoring this would make the interval too narrow (Donner & Klar, 2004). The estimated intraclass correlation among the twelve classes was 0.06.

The interval runs from a small effect to a large one, so the data are consistent with a class that adds only six points of completion as well as one that adds more than a third, and the agency's decision should rest on that whole range, not on the central figure alone.

Methodologists warn against treating an interval as a statement that the true effect sits inside it with 95 percent probability; it marks the effect sizes that fit the data well given the model's assumptions (Greenland et al., 2016). The lower end, six points, falls well below the 20-point gain the agency's director named as worth the class's staffing. The data therefore do not settle that question.

Loss to Follow-Up

The 55 adults lost by day 42 differed from those who stayed. Their baseline completion was lower, 26 compared with 30 percent, and more of them were men who had started overnight work. If lost adults would have scored lower at day 42, the retained figures overstate completion in both groups. The effect on the difference between groups is less certain, because losses were similar in size. Reasons recorded in the assessor log were mainly new jobs, moves out of the city and changed phone numbers.

A sensitivity analysis tested a pessimistic case for the class: lost early starters were counted as not completing the task at day 42, and lost waiting adults were assumed to keep their baseline result. Under those assumptions the difference falls to about 11 points. The true value likely lies between that figure and the main estimate.

Language Groups

Results by language are reported in Table 3 for description only. The Swahili subgroup is small, 9 early starters and 19 waiting adults, and supports no separate conclusion. For Dari and Arabic speakers, the direction of the difference matched the overall result. Adults using telephone interpreters showed the smallest gain in either group, consistent with the agency's concern that the class reaches them least well, but the numbers are too small to say more.

Secondary Measures

On the five-point item about filling out medical forms, mean confidence among early starters climbed from 2.4 to 3.9, compared with 2.5 to 3.1 among waiting adults. As Unit 4 expected, confidence rose more than ability, and it is reported only beside task results.

Among adults who consented to clinic release, 71 percent kept a first primary care visit within 90 days. Because waiting adults had also taken the class by day 90, this figure describes all class participants and is not a comparison.

The Second Scenario

Completion of the vaccination-record call, added by the council, rose from 22 to 51 percent among early starters and from 20 to 30 percent among waiting adults. That call was not part of the class's original content, and educators added a short practice of it only after the council's change, so the result is reported as exploratory. It suggests the class's phone skills carried over partly to a call it barely taught, which the agency may find as useful as the main result.

What the Interviews Explain

Eighteen interviews explained mismatches between task results and attendance. Adults who passed the call yet never reached a first visit described shift schedules that clashed with clinic hours and appointment letters sent only in English. Adults who did not complete the task but attended had usually had a relative or caseworker make the call for them. These themes explain patterns; they are not counted, since interviewees were chosen on purpose.

What the Findings Support

The design compares groups that were not formed by chance (Shadish et al., 2002), and the main threat, the language imbalance, was adjusted for but not removed. The findings support this statement: task completion improved more among early-cohort graduates than among comparable adults waiting for a seat, by an amount that could be modest or large.

References

Donner, A., & Klar, N. (2004). Pitfalls of and controversies in cluster randomization trials. American Journal of Public Health, 94(3), 416-422. https://doi.org/10.2105/AJPH.94.3.416

Greenland, S., Senn, S. J., Rothman, K. J., Carlin, J. B., Poole, C., Goodman, S. N., & Altman, D. G. (2016). Statistical tests, P values, confidence intervals, and power: A guide to misinterpretations. European Journal of Epidemiology, 31(4), 337-350. https://doi.org/10.1007/s10654-016-0149-3

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