12.7.1: Race and Depression Treatment

A significant amount of research has demonstrated disparities in health care treatment with respect to race in the United States. One study specifically looked at disparity in depression treatment among racial and ethnic minority populations in the United States (Alegría et al. 2008). The study is based on several surveys that focused on the collection of epidemiological data on mental disorders with a special emphasis on the usage of resources across different racial and ethnic groups. These surveys use a nationally representative sample. Psychiatric disorders were assessed using standard diagnostic measurement systems. 

All respondents to one of the surveys were asked questions about mental health services and use of prescription medication over the previous year. To define access to mental health care, the survey assessed whether the individual received any mental health treatment in the past year. The quality of depression treatment relied on a standard definition, which relies on the current professional knowledge of care.  

The reported statistical analyses from this study are quite numerous, so we will focus on only a few results. The first part of the analysis that we will consider is based on 8,762 individuals who responded to the Collaborative Psychiatric Epidemiology Survey. Some of the results from Table 1 of the research paper are given in Table 12.2. The results include the percentage of individuals who responded to the survey who had a depressive disorder during the past year. 

 

Table 12.2 Results from 8,762 individuals who responded to the Collaborative Psychiatric Epidemiology Survey. This table is adapted from the disorder category results show in Table 1 of Alegría et al. (2008). The results shown are the percentage of individuals experiencing any depressive order in the past year and the estimated standard error.

African Americans		Asians		Latinos		Non-Latino Whites
Percent	SE	Percent	SE	Percent	SE	Percent	SE
8.0	0.6	5.4	0.9	10.8	0.9	11.2	0.6

 

From the results one can observe that 11.2% of non-Latino whites reported that they had a depressive disorder in the past year. This is an example of an estimate in that this percentage is meant to give us an impression of what is happening in the entire population. The standard error is estimated to be 0.6. This means that the typical error in the estimate of 11.2 would be about 0.6, indicating that we might expect that the percentage could easily be as low as 10.6 or as high as 11.8 where we used one standard error on each side of the estimate. The standard errors can be used to make some crude comparisons, though some care must be used. Suppose we would like to know whether the percentage of Asian individuals experiencing a depressive episode is different than the percentage of African Americans in the population. These two percentages are estimated to be 5.4 and 8.0, respectively. While the two estimates are different, we must account for possible error. The standard error for the first percentage is estimated to be 0.9 while the standard error for the second percentage is estimated to be 0.6. To be safe, looking at two standard errors on each side of the percentage estimates we see that the percentage for Asians could be as low as 3.6 or as high as 7.2. For African Americans the percentage could be as low as 6.8 or as high as 9.2. Hence, we cannot really conclude from these results that the two percentages are different in the population because they are too close together. On the other hand, the percentage for Latinos is 11.2 with an estimated standard error equal to 0.6. Again, using two standard errors, this means that the percentage in the population could be as low as 10.0 or as high as 12.4. In this case it seems reasonable to conclude that the two percentages are probably different in the population. 

The next set of results are given in Table 12.3, which is adapted from Table 2 in the original research paper (Alegría et al. 2008). This table shows the quality of care received by individuals who had reported a depressive disorder across the racial and ethnic groups. Focusing on the first two columns of results, we can observe the percentage of individuals who had reported a depressive disorder and had received no treatment. There are clear differences between the percentages of those receiving no treatment across the racial categories. An analysis with the reported standard errors is possible here, but in this case the authors have provided results for a statistical test. In the footnotes to the table in the original paper, the authors state (Alegría et al. 2008):

Wald tests to identify differences in each of the treatment types across the racial and ethnic groups were significant at p<0.001. 

The first two words tell us that they used a statistical test called a Wald test. The specific type of test that is being used here is not important to our purpose here; we are interested in the null and alternative hypotheses, the conclusions, and the risk levels involved. The next part of the sentence tells us the test was used to identify differences in each of the treatment types across the racial and ethnic groups. The authors of the study tested whether the percentages of individuals receiving no treatment were the same between the racial or ethnic groups. Because the null hypothesis must have the concept of equality, the null hypothesis is that the percentage of individuals receiving no treatment is the same. The alternative hypothesis is what is true when this hypothesis is false. That is, there is at least one difference between these groups. Note that this does not mean every group will differ from all other groups. The alternative hypothesis means that there is at least one pair of groups that differ from one another. The next part of the sentence states that the results were significant, meaning that they rejected the null hypothesis and sided with the alternative hypothesis. No significance level is stated, but the final part of the sentence reports that the \(p\)-value is less than 0.001. This means that the null hypothesis would be rejected when \(\alpha=0.05\) and when \(\alpha=0.01\). The significance level could even be as small as 0.001, and the null hypothesis would still be rejected. This is a very strong conclusion stating the percentage of individuals who received no treatment for a depressive disorder are different by race and ethnicity. 

 

Table 12.3 Percentages of individuals withing each race or ethnicity who had reported a depressive disorder by quality of treatment. This table is adapted from Table 2 of Alegría et al. (2008).

	No Treatment		Inadequate Treatment		Adequate Treatment
Group	Percent	SE	Percent	SE	Percent	SE
Non-Latino White	40.2	3.1	26.8	2.8	33.0	1.90
Latino	63.7	4.5	13.9	2.5	22.3	4.9
Asian	68.7	6.8	18.1	5.0	13.1	4.6
African American	58.8	3.2	29.0	3.1	12.1	2.5

 

The final part of the analysis we will consider includes some confidence intervals. Table 12.4 shows the observed percentage of those who received adequate treatment for their depression within each race and ethnicity. The second column in the table shows the percentage from the sample that is being used to estimate the corresponding percentage in the population. The confidence intervals for each group are given in the third column. For non-Latino white individuals, the confidence interval starts at 28.9 and ends at 37.9. This means that there is 99% confidence that the true percentage of non-Latino white individuals with depression who receive adequate treatment is between 28.9% and 37.9%. Similarly, for Latino individuals there is 99% confidence that the true percentage of non-Latino white individuals with depression who receive adequate treatment is between 16.7% and 32.7%. Note that there is some overlap between these two confidence intervals, which provides evidence that the difference between the observed percentages could be simply due to estimation error. On the other hand, for African American individuals there is 99% confidence that the true percentage of African American white individuals with depression who receive adequate treatment is between 6.4% and 15.7%. This does not overlap with either of the intervals for non-Latino white and Latino individuals, so it appears that the percentage for African American individuals is lower. 

Table 12.4 Percentage and confidence intervals of the individuals with depression who received adequate treatment by race and ethnicity. This table is adapted from Table 4 of Alegría et al. (2008).

Racial or Ethnic Group	Sample Estimate	95% Confidence Interval
Non-Latino White	33.4	28.9 to 37.9
Latino	25.0	16.7 to 32.7
Asian	18.9	7.9 to 31.5
African American	10.4	64 to 15.7