The percentage of respondents scoring at the floor (the lowest score) or ceiling (the highest score) was acceptably low for all scales except first contactutilization, where 50% of the respondents scored the maximum score.
Table 6 compares the alpha coefficient and interfactor correlation for each primary care scale. The alpha coefficient of each scale substantially exceeded its correlation with all other primary care scales. None of the inter-factor correlations were excessively high, demonstrating that each primary care scale has significant unique contribution. All significant correlations were positive, indicating the complementary nature of primary care domains. Relatively high and positive interfactor correlations were observed between comprehensivenessservices received and comprehensiveness-services available (0.44), with the former and longitudinalityinterpersonal relationships (0.43), with the latter and coordination (0.38), and with comprehensivenessservices received and community orientation (0.37).
Discussion
Using patient-provided survey information collected within 2 health plans in South Carolina, we assessed the validity and reliability of the PCAT-AE. The results indicate that the hypothesized scales for primary care (first contactaccessibility, first con-tactutilization, longitudinalityinterpersonal relationships, comprehensivenessservices available, comprehensivenessservices received, and coordination) have substantial reliability and validity, consistent with the findings from the testing of the PCAT-CE.30 The 2 versions of the instrument differ only in the comprehensiveness domains, as comprehensiveness implies that all common needs are met, and health needs in childhood are different from those in adults. In contrast, challenges to accessibility, to the nature of interpersonal relationships, and to coordination and community orientation are similar for both children and adults and thus can be assessed by the same items. Only 1 ancillary feature of primary care, community orientation, was retained as a separate dimension after factor analyses. The extracted factors explained 88.1 percent of the total variance in the item scores.
All of the 5 assumptions, including item-conver-gent validity, item-discriminant validity, equal item variance, equal item-scale correlation, and score reliability, were met. These results suggest that these items may be used to represent the primary care scales, and the scoring of these items may be summed without standardization or weighting, as with Likerts method of summated rating scales.39
The resulting instrument has 74 items. Although the retained items adequately addressed first contactutilization, longitudinalityinterpersonal relationships, comprehensivenessservices available, comprehensivenessservices received, and coordination, and are consistent with the framework, those representing first contactaccessibility fell short. Only 4 of the 12 items measuring accessibility were retained. When more detail on accessibility is required, items that were deleted because they had lower item-total correlation may be added back in. Users should also review the comprehensiveness items to ascertain their relevance in the setting in which they are to be used. Items may be deleted if they are inappropriate in the context in which they are used; for example, in health systems that do not offer on-site testing for human immunodeficiency virus (HIV), because HIV is uncommon. Since continuity of care is an important component of primary care quality, a minimum number of visits or minimum duration with a regular source of care should be part of the assessment tool.
Separate factor analyses were performed with the 2 health plans. The results were largely comparable in terms of the factors that emerged as significant, indicating the generalizability of the tool to both vulnerable and middle-income populations. The only major differences are that the CHC subpopulation analysis yielded an additional significant factor, cultural competence, which the HMO subpopulation and the total population analyses failed to identify. In contrast, the HMO subpopulation analysis yielded an additional significant factor, family centeredness, which the CHC subpopulation and the total population analyses failed to identify. Thus, when using PCATon vulnerable populations (especially racial and ethnic minorities), questions measuring cultural competence might be retained. Family centeredness seemed to emerge as a distinct concept, primarily in the middle-income population.
There are a number of uses for a valid and reliable instrument such as the PCAT-AE. First, understanding primary care as a multidimensional concept is consistent with the IOMs conceptualization of primary care and more precisely captures the quality of primary care than unidimensional proxies, such as a clinicians medical specialty. With the 6 scales representing 4 core domains, the index representing strength of affiliation with a primary care provider, a scale for community orientation and the optional scales for family centeredness and cultural competence, all the important features of primary care are addressed. Second, PCAT-AE can be used as a quality measurement tool that assesses the adequacy of primary care experience rendered under different health care systems or settings, and for patients with different sociodemographic attributes. Third, PCAT-AE can also serve as a quality control tool that compares the quality of primary care given by providers of different types. The instrument can be used with other outcomes to assess the effect of policy interventions and systems changes on the delivery of critical aspects of primary care.