This study is one of the few to test multiple SES measures on a critical health-related outcome in the pediatric population and since SES measures for both patient and parent were based upon the parent’s responses to SES items, these can be considered valid measures of the family’s socio-economic status. Throughout the study, family income demonstrated the strongest fidelity in predicting HRQOL scores, while the impact of parental educational attainment level and the Hollingshead Index on HRQOL scores was somewhat muted. For the patient models, family income (η2 = 0.039) had 3 times higher the amount of variance explained compared to either the Hollingshead Index or parental educational attainment level. In the parent model, parental educational attainment level increased the η2 by 0.024 when controlling for covariates; however, family income by itself significantly increased the amount of variance explained by more than twice that amount, (η2 by 0.050). When examining the final model with parental educational attainment level and family income, since parental educational attainment level was not significant, while family income remained significantly related to parent-reported PCQLI score, we can conclude that family income by itself shows the strongest relationship with both patient- and parent-reported PCQLI Total score. While 4-5% of the variance explained by family income may seem small, the correlation coefficient associated with an η2 of that size is approximately 0.20, indicating a low moderate effect size . Even though this effect may not have much clinical relevance for individual patients, for large populations, understanding the influence of family income could inform the level of risk or protection for health related quality of life based on SES, over and beyond health-related factors.
Our results regarding family income as a predictor of HRQOL were consistent with other studies that examined the relationship between income and health status. Even though this study did not translate family income into percent Federal Poverty Level (FPL), our findings were similar to the Newacheck’s  study on health care disparities among adolescents. They found strong gradient differentials between income (FPL) and general health status. Additionally, our results seem to confirm the existence of social gradients and aspects of health in adolescents found in Starfield’s study .
For many, family income is perceived as only associated with problems of health care access and utilization, and these problems are currently being addressed by programs such as Medicaid and the State Children’s Health Insurance Program (SCHIP). While these associations are true, other income-related issues are just as salient. For example, income levels are positively associated with better “nutrition, housing, schooling, and recreation,”  all elements relevant to an individual’s HRQOL. Individuals either living in poverty or near the poverty line are more likely to have problems with access to health care, have lower rates of health care utilization, and report that they have less satisfaction with care than individuals with higher SES scores [24, 27, 28]. Furthermore these SES factors have a deleterious effect in that individuals at the lower ends of the SES spectrum have higher rates of morbidity and mortality [29, 30]. Fewer financial barriers to accessing health services might result in higher rates of preventative care and ultimately a healthier population.
Parental education as a proxy for SES, as suggested by Winkleby , did not yield as much predictive power on HRQOL as family income. As previously stated, while income and elements of HRQOL are associated, parental education is related to other aspects of a child’s health. For example parental education has been shown to affect child cognitive development  specifically but may only be broadly associated with HRQOL.
Finally, many researchers might be hesitant to ask income questions for their projects, relying on the Hollingshead Index instead. The Hollingshead Index requires two pieces of information: 1) parental educational attainment level and 2) occupation. On theoretical terms, the occupational component is much more problematic. Relying on a job title as an indicator of a person’s position within the social structure without any knowledge of their potential income may not be the best way to measure SES. Statistically, the Hollingshead Index had the lowest predictive power of the three indices. If researchers are going to take the time to ask sensitive questions about socio-economic status, results from this study indicate that income can be used in studies even if the categories have a broad bandwidth. These broader categories may allow respondents to be comfortable enough to answer the income question without feeling as though the question is intrusive. More importantly, however, is the need for researchers to re-examine the use of the Hollingshead Index as a measure of SES.
Further studies are needed to test the individual-level income categories against other measures of SES recently developed with the use of geocoding on HRQOL instruments, including the PCQLI. Although there have been significant limitations in using community-level geocodes for individual-level data [33, 34], others have noted positive results from the use of geocodes [35, 36]. If an SES measure can easily be developed with the use of geocodes and have as strong a predictive value as family income, researchers may have another SES tool with which to work. A second recommendation would be using, whenever possible, multiple measures of SES  Adler suggests using multiple determinants of SES to examine, and ultimately eliminate or reduce, disparities in health . With parental education and family income having a moderate correlation, these two SES measures are capturing different aspects of socio-economic status. Parental education is a simple variable to collect, and as shown in this study, most participants are willing to provide family income information if the income categories are broad enough.
This study only had a single time point of analysis and thus cannot show change over time. Additionally the income distribution for the sample does not reflect the general population of the United States. Other limitations include the potential for bias due to the exclusionary criterion of English-speaking only.