Clinicians and researchers frequently use the Medical Outcomes Study’s 12-Item Short-Form Survey Instrument (SF-12) to assess and monitor health-related quality of life. (The term "item(s)" henceforth will refer to SF-12 questions with the associated response categories.) The SF-12 measures eight attributes of functional health status: physical functioning, role limitations resulting from physical health problems, bodily pain, general health, vitality (energy and fatigue), social functioning, role limitations resulting from emotional problems, and mental health (psychological distress and psychological well-being). In addition, the SF-12 assesses overall physical and mental function using summary scales, Physical Component Summary Score (PCS-12) and Mental Component Summary Score (MCS-12), which are scored through comparison with population norms estimated from responses to the 1990 National Survey of Functional Health Status (1,2).

The SF-12 has been used to measure the health status of patients with a specific diagnosis (3-9) as well as that of the general population, including health plan enrollees and various age and ethnic groups (10-13). The survey has been administered by telephone (4,14), self-administered in a clinic setting (10,14,15), administered by personal interview in a clinic setting (8), and administered as a mail survey (14,16,17). There is growing interest and some evidence of validity for using the SF-12 as a measure of population health status for purposes of planning, implementing, and evaluating community health interventions (16). Although research has shown the SF-12 to be valid for specific demographic groups (10,11,14,18-20), few studies have examined the psychometric properties of the SF-12 when used with minority populations in the United States (15,17,21-22). The dearth of research regarding the use of the SF-12 for American minority populations underscores the importance of further research because it was developed and validated in an ethnically homogeneous, predominantly white population. As with any measure of health related to quality of life, it is important to understand the psychometric characteristics to assure that appropriate interpretation and adequate inferences are made about the population of interest (23).

In light of alarming disparities in the health of the disadvantaged, including minority groups and those with low incomes, community-based participatory research (CBPR) initiatives have become more prevalent as a means of improving population health, often replacing more conventional approaches. CBPR is a partnership approach that involves community members, researchers, and organization representatives in all stages of the research process, including identification of key issues, study design, development and implementation of interventions, evaluation, and dissemination of research findings. This approach is often developed and implemented through community coalitions and relies on decision-making processes that empower all partners to contribute their knowledge and expertise, thus facilitating a collaborative, equitable relationship with shared ownership and responsibilities. The basic principles of CBPR are community member engagement, using local knowledge in the design of interventions and investing community members in the process and products (24). Trust between researchers and community partners is critical, albeit a challenge in light of past research experience in which the community derived little or no benefit or feedback from study results (25). Research studies should employ research tools, surveys, and questionnaires that are sensitive to the culture and norms of the population, to its race/ethnicity, age, social class, language, reading level, or religious customs (25). Such research tools not only will ensure valid results but also will build confidence and trust in academic and community partnerships. In addition, if evidence of validity is demonstrated, measures such as the SF-12 or other health-related functional health indicators can be recognized as useful health status measurement tools to inform the effectiveness of community-based health interventions with specific racial/ethnic populations.

The purpose of our study was to assess the validity of the SF-12 for establishing a baseline for mental and physical health of African Americans within the context of a CPBR study, Nashville's Racial and Ethnic Approaches to Community Health (REACH) 2010 initiative, which seeks to reduce disparities in diabetes and cardiovascular disease between African Americans and whites (26). The analytic approaches presented here test the validity and reliability of the SF-12 instrument within the context of a larger community-based data collection effort.

A total of 1721 African American adults responded to our survey. Of these, 61% were female with an average age of 53.2 years (Table 1). Fifty-three percent reported at least a high school education, and 52.8% were employed. Comparison of this sample with U.S. Census 2000 demographics shows that the sample underrepresented younger adults (aged 18–24 years) and educated adults (post-high school education) and overrepresented older adults (aged ≥65 years).

As seen in Table 2, 11% of respondents reported having been previously diagnosed with diabetes, which is considerably higher than the U.S. average of 7% (32). In addition, based on self-reported height and weight, 31% of this sample was obese. Twenty-four percent reported their overall health as fair or poor (based on a five-point Likert scale [excellent, very good, good, fair, poor]). Thirty-five percent reported that they had had 1 day or more during the past month when their physical health was not good, whereas 23% reported that they had had 1 day or more during the past month when their mental health was not good. Our indicator of social support, measured by the number of friends or relatives who were available to offer emotional support when needed, showed that only 3% of the sample had no one available for help with emotional problems.

Based on the results of this psychometric evaluation, SF-12 summary scores were reliable and valid for use with African Americans as measured within the context of a larger community-based study. At the item level, very good data completeness was obtained, demonstrating a willingness of this sample of adults to provide information about their health that bears well on the validity of the scale scores because loss of items from nonresponse bias is minimal. Ceiling effects were observed on several items, suggesting that these items may not be sufficiently sensitive to change for longitudinal use.

The two items that assessed the effects that physical health had on role functioning (Accomplished less than you would like [Q4]; Were limited in the kind of work or other activities [Q5]) and the two items that assessed the effects that emotional health had on role functioning (Accomplished less than you would like [Q6]; Didn't do work or other activities as carefully as usual [Q7]), as well as the social functioning item (How much of the time has physical or emotional problems interfered with your social activities [like visiting with friends or relatives] [Q12]), demonstrated that more than 75% of respondents had no impairment. As suggested by McHorney et al (34), it may be beneficial to replace the dichotomous response categories (i.e., presence or absence of limitations) with finer gradation of response categories. These results suggest that responses to these questions may not accurately represent the true level of role functioning of this sample. This may also be an artifact of the telephone mode of administration, in which respondents may minimize impairment of functioning in these areas when reporting verbally. Additionally, because the SF-12 was embedded in a longer questionnaire, the questions preceding the SF-12 items could have influenced SF12 responses to be more positive than if respondents were administered only the SF-12 survey. The instrument should be further tested at the item level among this population.

At the scale level, the MCS12 and PCS12 showed satisfactory reliability in internal consistency. In spite of the limitations at the item level, the scale scores did not show ceiling or floor effects that would support their use longitudinally. Construct validity was supported for the physical health summary score by differentiating between those individuals who would be expected to have a lower level of physical functioning based on being of older age, having diabetes, or being obese. Construct validity was also supported for the mental health summary score by differentiating between those who would be expected to have a lower level of mental health functioning based on the presence or absence of social support in their lives. Convergent and discriminant validity was evidenced by satisfactory intercorrelations between items and summary scale scores consistent with the construct. Items that pertained to emotional functioning, mental health, and social support were more highly correlated with MCS12 than with PCS12. Likewise, items that pertained to physical health, physical role functioning, general health, and bodily pain were more highly correlated with PCS12 than with MCS12. As might be expected, vitality was correlated fairly equally with both summary scores.

Concurrent validity was supported by the strong correlation of the MCS12 and PCS12 with the number of days respondents reported feeling in poor mental and physical health during the past month. These results show that the summary scores can be used to describe functional limitations in physical and mental health.

The physical health summary scores were marginally lower than the U.S. general population norms, consistent with the percentage of respondents who were older, had diabetes, or were obese. Interestingly, the mental health summary scores were found to be higher for this African American sample compared with the national norms of the U.S. general population. This may be an artifact of the item-level skewness observed for three of the five items that the MCS12 comprises, although ceiling effects were not observed at the scale level. It is possible that a telephone mode of administration elevated the MCS12 scores. Participants may have given more positive reports as a result of demand characteristics or social desirability when speaking with a telephone interviewer as opposed to completing an anonymous mail survey or speaking with a clinician. McHorney et al (35) determined that scores are higher when health status is reported by telephone compared with reporting health status by mail. Further research is needed to understand this potential effect.

We acknowledge several limitations of this study. The telephone methodology used to collect these data imposes limitations associated with noncoverage and nonresponse bias. A telephone survey methodology that employs a list of residential telephone subscribers has noncoverage bias against people without telephones, although the U.S. Census 2000 reports 94.5% telephone coverage in the geographic area we targeted (27). This methodology also biases against people who have unlisted telephone numbers and people who can be reached only by cell phone. Noncoverage bias could be reduced by combining a stratified residential telephone list method with a random-digit–dialing method that would include people who have unlisted numbers.

Nonresponse bias was evidenced by the low response rate (34.9%) and a high refusal rate (23%). This is consistent with declining response rates in national telephone surveys over the past several years (37). Also, households within highly urbanized areas of concentrated socioeconomic disadvantage are less likely to participate in telephone surveys (38).

Several additional factors may have contributed to the refusal rate within this study. The potential respondents were told the interview would take 30 to 40 minutes. This may have contributed to refusals among participants, particularly the young adult age group. Young adults (aged <25 years) were the most underrepresented in our study consistent with research that has shown that adults in this age group are among the hardest to reach in telephone surveys (36). Another contributing factor may have been that no monetary incentive or other compensation was offered for respondents' time to complete the survey. Finally, there may exist subgroups within our population of interest, such as Caribbean Islanders, Africans, and people of other origins, who may have been less likely to respond to telephone surveys as a result of language barriers or cultural influences.

A mail methodology or combination mail and telephone methodology may achieve higher response rates among low-income African Americans. Previous research has shown higher response rates to the SF-36 survey using a mail methodology than with a telephone methodology (35). Within a Medicaid population, mixed methodology surveys (user-friendly mailed surveys) with incentives (37), combined with telephone follow-up, have been shown to increase response rates (39). Incentives have been shown to enhance response rates among African Americans (40).

Further research could explore alternative and mixed methodologies within a low-income African American community. If administration of the SF-12 is embedded within a longer community survey, response rates may be less compromised by limiting the number of additional questions to keep the interview or mail survey as brief as possible. In addition, offering respondents incentives for survey completion may increase response rates regardless of mode of administration.

Collectively, results of our study demonstrate validity for cross-sectional use of SF-12 summary scales for measuring the health status of low-income African Americans. In addition, this study shows how simple procedures can be used to assess measurement properties of the survey instrument when used in a population that was not adequately represented in the original studies of the SF-12 instrument. Within a CPBR initiative, this approach will provide assurances to community members that will increase acceptance of the study results and will increase trust in academic–community partnerships by using or testing instruments that are culturally appropriate. It is imperative that future validity studies of new and existing instruments recruit African American participants as well as members of other racial and ethnic groups. We will focus our future work on the use of the SF-12 in longitudinal studies of health among low-income African Americans.