A Community Navigation Scale for Persons With Serious Mental Illnesses

The Community Navigation Scale is a 21-item scale that was created to address a gap in outcome measurement for two studies of the ODR model (15, manuscript submitted). Two investigators (B.B., M.T.C.) designed items on the basis of the four ODR tenets: ensuring adequate treatment (e.g., getting my medicine), helping to secure safe and stable housing (e.g., keeping safe and stable housing), developing a meaningful day (e.g., being involved in the community and volunteering in the community), and using technology to promote recovery (e.g., using a computer and using a cell phone). Additional items were developed to assess three other domains of community functioning and psychosocial well-being, including basic activities of daily living (access to reliable transportation, managing money, shopping for groceries, availability of small appliances, cooking meals, keeping the house clean, and keeping up daily hygiene and grooming), positive health behaviors and protective factors (regular exercise schedule, eating a healthy diet, forming social relationships, receiving needed support, and having a satisfying spiritual life), and navigating one’s community (finding classes to learn new things, finding enjoyable activities, and knowing where to go for help finding a job). Some items tapped several of these overlapping domains (e.g., “I volunteered in the community” pertains to both a meaningful day and navigating one’s community).

Domains and items were reviewed by several subject matter experts during an iterative process of item development, which resulted in the 21 items. Items were then further revised on the basis of a review by three mental health services researchers, four licensed mental health professionals working as professional navigators on the initial ODR implementation (i.e., licensed clinical social workers and licensed professional counselors), and four certified peer specialists working as peer navigators on the initial ODR implementation. Although each item was initially scored on an evenly spaced, 7-point Likert scale ranging from 1 (“very hard”) to 7 (“very easy”), consensus discussions with reviewers resulted in changing the end points of some items according to the content of those items while keeping the 7-point Likert scale. Specifically, scales ranged from “very hard” to “very easy” (13 items), “very untrue” to “very true” (four items), “very unimportant” to “very important” (three items), or “very unsatisfying” to “very satisfying” (one item) according to what was felt to be most appropriate and understandable to the respondent. Pilot testing was conducted before the Community Navigation Scale was used. Specifically, the instrument was used with 10 patients, with clinicians giving feedback on wording and patients’ comprehension of items; very minor changes were made in finalizing the scale.

Both research projects also used the MCAS, a 17-item self-report instrument that examines several dimensions of community functioning. The scale’s validity is good, and test-retest reliability and internal consistency are high (16). The internal consistency reliability of the MCAS in this combined sample was α=0.85.

During the initial ODR study (15), 100 participants were enrolled in the study before being discharged from a local state psychiatric hospital or one of three crisis stabilization units. Inclusion criteria for participation were ages 18–65 years, English speaking, diagnosed as having a psychotic or mood disorder, being discharged to reside within the catchment area of one of the three community service boards, and ability to give informed consent. Exclusion criteria were known or suspected intellectual disability or dementia, or a serious medical condition that could interfere with research participation.

The second study enrolled 240 participants in a randomized trial comparing ODR with traditional forms of case management; participants were enrolled before being discharged from the same state psychiatric hospital or one of two crisis stabilization units. The two samples were assessed at different times given the studies’ consecutive nature. Inclusion criteria were nearly identical, but the second study had stricter functional impairment requirements; specifically, because of the randomized design, participants in the second study had to be eligible for intensive case management services in Georgia.

Trained research assessors conducted interviews with the participants. Each interview lasted approximately 2–3 hours, with the Community Navigation Scale assessment taking 5–7 minutes. These instructions were read at the beginning of the instrument: “Now I’ll read several statements to you. For each statement, I would like you to choose the answer that best fits your opinion.” Each item began with “During the month before you came to the hospital or CSU [crisis stabilization unit]. . . .” The participant was shown a response card with the respective Likert scale. Participants were compensated with $80 for the entire interview. Both studies were approved by the institutional review boards of The George Washington University (for the first study) and New York State Psychiatric Institute at Columbia University (for the second study).

We used baseline data for this analysis. Data from the two studies were compared and then combined for subsequent analyses because factor analysis is considered a large-sample technique. The Kaiser-Meyer-Olkin measure of sampling adequacy and Bartlett’s test of sphericity were computed to determine suitability of the data for factor analysis. The principal axis factoring method, followed by a varimax rotation, was used for exploratory factor analysis, and its results are given here. Our a priori convention for assigning items to factors and subscales was that only the largest loading for each item would be considered if it was greater than the a priori 0.4 loading threshold. This method is in line with common practice because 0.4 is generally considered a moderately strong loading; moreover, the practical usefulness of coefficients has often been judged to lie in the |0.30| to |0.40| range (17). As a secondary approach, oblique rotation was used. In doing so, overall factor structure results did not change. However, some factor loadings decreased, with some dropping below our a priori threshold of 0.4. Only participants with a fully completed Community Navigation Scale were included in the factor analyses (N=328). Subscale scores were calculated by adding item responses, and correlations between the scores were computed. We assessed the derived subscales’ internal consistency reliability with Cronbach’s alpha.

As an initial assessment of convergent validity, we examined the correlation between each Community Navigation Scale subscale score and the MCAS. We chose the MCAS because it is a widely used measure of community functioning among individuals with serious mental illnesses, measuring constructs somewhat similar to those of the Community Navigation Scale (e.g., some items pertain to critical abilities for coping with a serious mental illness and surviving in the community, including successfully managing money and managing day-to-day tasks such as eating regularly, dressing appropriately, or keeping up one’s home). All analyses were conducted with IBM SPSS Statistics, version 26.

The demographic and clinical characteristics of the study participants are given in Table 1. The possible range of the total 21-item Community Navigation Scale score was 21–147, the observed range was 23–146, the mean±SD was 89.7±24.1 (median=90, mode=105), and the Cronbach’s alpha coefficient indicating internal consistency reliability was 0.89. Distributions of responses across the 7-point scales were carefully examined for each of the 21 items and generally showed very good dispersion of scores across the scale. However, two items (6 and 9) showed right skewness. For item 6, “Keeping my house clean was...,” the following response frequencies (N=334 participants) were found: 1 (very unimportant), N=14 (4%); 2 (unimportant), N=18 (5%); 3 (a little unimportant), N=17 (5%); 4 (neutral), N=35 (10%); 5 (a little important), N=29 (9%); 6 (important), N=74 (22%); and 7 (very important), N=147 (44%). For item 9, “Keeping up my daily hygiene and grooming was...,” the following results (N=335 participants) were found: 1 (very unimportant), N=4 (1%); 2 (unimportant), N=12 (4%); 3 (a little unimportant), N=18 (5%); 4 (neutral), N=25 (7%); 5 (a little important), N=15 (4%); 6 (important), N=80 (24%); and 7 (very important), N=181 (54%).

The Kaiser-Meyer-Olkin measure of sampling adequacy was 0.90, and Bartlett’s test of sphericity was statistically significant (χ²=2,231.37, df=210, p<0.001), indicating that the data were suitable for factor analysis. The initial principal-axis factor analysis revealed four factors with eigenvalues >1.0, which accounted for 52% of the total variance. A thorough examination of eigenvalues and the cumulative proportion of explained variance, as well as visual inspection of the scree plot showing ordered eigenvalues by factors, suggested reducing the number of factors to three for further analysis. These three factors had eigenvalues of 6.58, 1.85, and 1.50 and accounted for 47% of the total variance. After varimax rotation, the factor solution uncovered the factor loadings presented in Table 2. As noted previously, our a priori convention was that only the largest loading for each item, if >0.4, would be considered in assigning items to factors.

Factors to which items were assigned are shown in Table 2. Factor 1 was labeled “social and physical well-being,” factor 2 was “accessing external resources,” and factor 3 was “home and self-maintenance.” The three subscales’ distributional properties were examined. Although social and physical well-being and accessing external resources did not exhibit meaningful skewness (−0.003 and −0.057, respectively), home and self-maintenance was slightly skewed (−0.602). Kurtosis values (−0.675, −0.663, and −0.323, respectively) did not indicate outliers or a need to consider the distributions as substantially nonnormal. Cronbach’s alpha indicated good internal consistency (Table 3). Means, standard deviations, possible ranges, observed ranges, and medians for the three subscales, along with intercorrelations, are reported in Table 3.

Concerning convergent validity with the MCAS total score, medium-strength correlations that were statistically significant at the p<0.001 level were found for all three factors (social and physical well-being, r=0.65; accessing external resources, r=0.55; and home and self-maintenance, r=0.41) and the Community Navigation Scale total score (r=0.66).