Errors appeared in this article by Flores and Tomany-Korman (doi: 10.1542/peds.2007-1243). The publicly available 2003 National Survey of Children's Health (NSCH) dataset was analyzed for this paper. Although publicly available NSCH data provide nationally representative estimates for most racial/ethnic groups, they do not do so for Asian/Pacific Islander and Native American children, as Asian/Pacific Islander race was not reported for children living in 46 states, and Native American race was not reported for 44 states.
To rectify these errors, analyses for Asian/Pacific Islander and Native American children were re-run using the non-public NSCH dataset, which does provide nationally representative estimates for both groups. These revised analyses did not alter any of the major study conclusions; a brief textual summary of changes is provided below for each table. The revised tables (listing only those changes from the original tables) are available from the authors upon request.
Table 1: Estimates changed by 0.2-1.6, and only one, non-significant P-value changed (for gender), from 0.62 to 0.54.
Table 2: The estimates changed by 0-6.4. Two P-values changed from non-significant to significant: hearing/vision problems and ≥3 ear infections in last 12 months. The P-value for diabetes changed from significant (.04) to marginally non-significant (.05), and two P-values changed slightly but remained significant (headaches and speech problems).
Table 3: The estimates changed by 0-8.4, except for reasons for unmet medical care need, which changed by 0-58. The P-values for transportation barrier, treatment is ongoing, and health plan problem changed from significant to non-significant (.06), and three P-values changed slightly but remained significant (no insurance for both unmet medical and dental needs, and doctor didn't know how to provide care).
Table 4: The estimates changed by 0-3.1, and there were no changes in any P-value.
Table 5-7: Only odds ratios (ORs) and 95% confidence intervals (CIs) that changed are presented. For Latinos, there were only minor changes in a few ORs, except that significant odds were found for no routine preventive dental visit in the past 12 months. For African-Americans, there were only minor changes in a few ORs, except that the OR no longer was significant for emotional, developmental, or behavioral problems needing treatment or counseling. For Asians/Pacific Islanders, there were only minor changes in a few ORs, except that several not-estimable cells converted to not significant (for reason for unmet medical care need), the OR for teeth condition not excellent/very good became significant, and the OR no longer was significant for did not receive all needed medical care. For Native Americans, there were only minor changes in a few ORs, except that four ORs changed from non-significant to significant (has limited abilities, asthmatic, skin allergies, and any problem getting specialty care), four ORs changed from significant to non-significant (behavior problems, transportation as a reason for unmet medical needs, and dentist didn't know how to provide care and no dental insurance as reasons for unmet dental needs), and an OR incorrectly listed as significant in the prior version is now correctly listed as non-significant (for needs/uses prescription medication). For multiracial children, there were only minor changes in a few ORs.