PHLT 8032 Week 11 Final Analysis Report Example

Health Insurance, Income and Self-Reported Health Among Adults in a Composite County Survey: A Final Analysis Report Assembling Ten Weeks of Preparation, Description and Inference in SPSS

Student Name

Walden University

PHLT 8032: SPSS Revealed

Instructor Name

Month Day, Year

Health Insurance, Income and Self-Reported Health Among Adults in a Composite County Survey: A Final Analysis Report Assembling Ten Weeks of Preparation, Description and Inference in SPSS

Introduction

County health departments track self-reported health because it is cheap to collect and carries real information: a single rating of general health predicts later mortality even after other risk factors are taken into account (DeSalvo et al., 2006), and counts of days of poor physical and mental health give a finer picture of the burden people carry from month to month (Moriarty et al., 2003). Two characteristics a department can act on, at least in part, are health insurance coverage and household income. Coverage determines whether residents can reach preventive and mental health care, and income shapes housing, food and stress. Randomized evidence from Oregon found that gaining Medicaid reduced the rate of depression among low-income adults within two years (Baicker et al., 2013), and a review of the quasi-experimental literature found that gaining coverage widens access to care, protects household finances and improves several measured health outcomes (Sommers et al., 2017).

This report asks two questions of a composite county adult health survey. First, do uninsured adults report more days of poor mental health and worse general health than insured adults? Second, how does self-rated health vary across household income brackets? It assembles the preparation, description and analysis carried out over the course into one account, with the data preparation record, the audited syntax and selected output in the appendices.

Methods

Data and Unit of Analysis

The analysis draws on the 2024 round of a composite county health survey of adults aged 18 and older, the Harlow County survey created for instruction. The file came in as 1,846 spreadsheet rows across 16 variables, one row per interviewed adult. Its final weight rebalances respondents toward the makeup of county adults by age band, sex and district; it does not include strata or cluster identifiers.

Preparation

Preparation followed a dated cleaning log (Appendix A). Its consequential decisions were these. One interview keyed twice was removed, leaving 1,845 respondents. Where the codebook used 88 to mean no bad days at all, that code became zero, while don't know and refused became missing values, which leaves 30 people without a usable mental health count and 26 without a physical one. Three heights entered in feet were set missing, two entered in centimeters were converted, and one implausible weight was set missing; 1,662 respondents have a valid body mass index. Three weekly activity totals above 4,200 minutes were set missing under a rule fixed before inspection. In all, 164 respondents (8.9%) declined the income question, and declining was commoner from age 65 (12.4% against 7.6% below that age). No values were imputed. A full rerun of the syntax from the raw spreadsheet reproduced every reported figure after three errors in earlier draft figures were corrected (Appendix B).

Measures

Insurance status is any current health insurance, yes or no. Self-rated general health has five ordered answers, with code 1 meaning excellent and code 5 meaning poor; it is analyzed as ordinal and summarized as the share reporting fair or poor health. Poor mental health days form a count bounded by the 30-day recall window and are described by medians, interquartile ranges and the proportion at 14 days or above, the usual cutoff for frequent mental distress. Household income comes in five ordered bands, the lowest under $25,000 and the highest $100,000 or more.

Analysis

Poor mental health days were compared by insurance status with the Mann-Whitney U test, chosen because the outcome is bounded and piled at zero, with r as the effect size. Fair or poor health was compared by insurance status with a Pearson chi-square test and phi. Income and self-rated health were related through percentages within each income band and Kendall's tau-b, chosen for its handling of the many ties two short ordinal scales produce (Agresti, 2010). Comparisons among respondents are unweighted. Estimates meant to describe the county's adults, shown in Figure 1, are weighted, with confidence intervals from the Complex Samples procedures using the weight alone; applying the weight through Weight Cases would have changed the percentages but understated their uncertainty (Lumley, 2010). Analyses used IBM SPSS Statistics.

Results

Respondents

Table 1 condenses the fuller profile built earlier in the term. Of 1,835 respondents with known insurance status, 213 (11.6%) were uninsured. Uninsured respondents were younger, more likely to smoke and more likely to report a household income under $25,000.

Table 1

Respondent Profile, Insured and Uninsured Adults Compared

Note. Unweighted valid percentages. Activity minutes and district appear in the full table in Appendix C.

Insurance and Self-Reported Health

Respondents without coverage tended to count more bad mental health days (z = -4.87, U = 138,050; p < .001; r = .11). Put another way, pick one adult from each group at random and the one lacking coverage reports the higher count roughly three times in five. Medians were 2 days without coverage and 0 days with it, and frequent mental distress affected 19.1% of the former against 9.8% of the latter.

Fair or poor general health was also more common among the uninsured: 26.3% of uninsured respondents (56 of 213) compared with 15.0% of insured respondents (243 of 1,616), chi-square(1, N = 1,829) = 17.5, p < .001, phi = .10. Both differences are small by conventional standards, though a near doubling of frequent mental distress is not small for a health department planning services.

Income and Self-Rated Health

Self-rated health improved steadily across income brackets (N = 1,675; tau-b = -.24, p < .001); the negative sign reflects the coding of health from 1 for excellent, so higher income went with more favorable ratings. Among respondents, the share reporting fair or poor health fell from 31.7% in the lowest bracket to 7.1% in the highest. Figure 1 shows the same gradient as weighted estimates for the county's adults.

Figure 1

County Adults Rating Their Health Fair or Poor Across Income Bands, Weighted Estimates

Note. Estimates use the final survey weight, which matches respondents to county adults on age band, sex and district (n = 1,675 respondents). Error bars show 95% confidence intervals computed with the weight alone; the extract carries no strata or cluster identifiers, so the intervals may still be somewhat narrow.

Discussion

Three findings stand out. Uninsured respondents carried more days of poor mental health and were more likely to rate their health fair or poor. Self-rated health followed a clear income gradient, with the share in fair or poor health more than four times as high in the lowest bracket as in the highest. And the uninsured were concentrated among younger, lower-income and smoking respondents, which complicates any reading of the insurance differences.

The direction of these patterns matches what national surveillance shows in general terms: fair or poor health is more common at lower incomes and among the uninsured in the Behavioral Risk Factor Surveillance System, whose self-reported measures have acceptable reliability and validity (Pierannunzi et al., 2013). Direct numerical comparison with published county or state figures would be misleading, however. The extract is a composite built for teaching, its year and definitions are not those of any published series, and most of the figures here are unweighted descriptions of respondents.

The insurance findings should be read with particular care. Because the uninsured are poorer and more likely to smoke, and because income is itself strongly associated with self-rated health, the unadjusted differences mix any effect of coverage with the effects of these characteristics. The direction could also run from health to coverage, since poor mental health can disrupt the employment through which many adults are insured. Randomized evidence suggests coverage does improve mental health (Baicker et al., 2013), but a cross-sectional survey cannot show it.

Limits

Five limits apply. All measures are self-reported, and body measures in particular are known to be biased in self-report. The survey is cross-sectional, so no finding establishes the order of cause and effect. No analysis adjusted for age, income or other characteristics; multivariable modeling was outside the course's scope and would be the next step. Complete-case analysis was used throughout, and income refusal was more common among older adults, so income results describe them less fully. Finally, the weighted estimates use the weight without the survey's design information, which likely understates their uncertainty.

Conclusion

In this composite county survey, lacking health insurance went with more days of poor mental health and a higher share in fair or poor health, and self-rated health worsened steadily as household income fell. The patterns are consistent with published evidence and with a case for attending to coverage and income together, but the unadjusted, cross-sectional analysis cannot separate them. The appendices hold the cleaning log (Appendix A), the audited syntax that reproduces every figure from the raw spreadsheet (Appendix B), and the full characteristics table with selected output (Appendix C).

References

Agresti, A. (2010). Analysis of ordinal categorical data (2nd ed.). Wiley.

Baicker, K., Taubman, S. L., Allen, H. L., Bernstein, M., Gruber, J. H., Newhouse, J. P., Schneider, E. C., Wright, B. J., Zaslavsky, A. M., & Finkelstein, A. N. (2013). The Oregon experiment: Effects of Medicaid on clinical outcomes. The New England Journal of Medicine, 368(18), 1713-1722. https://doi.org/10.1056/NEJMsa1212321

DeSalvo, K. B., Bloser, N., Reynolds, K., He, J., & Muntner, P. (2006). Mortality prediction with a single general self-rated health question: A meta-analysis. Journal of General Internal Medicine, 21(3), 267-275. https://doi.org/10.1111/j.1525-1497.2005.00291.x

Lumley, T. (2010). Complex surveys: A guide to analysis using R. Wiley.

Moriarty, D. G., Zack, M. M., & Kobau, R. (2003). The Centers for Disease Control and Prevention's Healthy Days Measures: Population tracking of perceived physical and mental health over time. Health and Quality of Life Outcomes, 1, Article 37. https://doi.org/10.1186/1477-7525-1-37

Pierannunzi, C., Hu, S. S., & Balluz, L. (2013). A systematic review of publications assessing reliability and validity of the Behavioral Risk Factor Surveillance System (BRFSS), 2004-2011. BMC Medical Research Methodology, 13, Article 49. https://doi.org/10.1186/1471-2288-13-49

Sommers, B. D., Gawande, A. A., & Baicker, K. (2017). Health insurance coverage and health: What the recent evidence tells us. The New England Journal of Medicine, 377(6), 586-593. https://doi.org/10.1056/NEJMsb1706645