Relational Coordination
Guidelines for Theory, Measurement and Analysis
Chapter 3: Measuring Relational Coordination (continued)
Administering the Survey
The relational coordination survey can be administered in person, by mail or by email. Once you have identified the functional groups you will survey, you need to survey participants from each functional group. For my flight departure study as a graduate student, I used a highly time intensive approach, delivering the survey in person, and being present to answer questions from respondents as they completed the study. At each of the nine participating sites, I administered the survey in person on a single day to employees working the morning shift, distributing surveys in the break rooms. All surveys were conducted on weekdays between Tuesday and Thursday, to avoid disrupting the operations and to increase the number of surveys completed because passenger loads were typically lighter on these three days. Respondents typically required 20 minutes to complete the survey. Four hundred surveys were administered with 354 completed, for an overall response rate of 89%.
For my patient care coordination study, as a junior faculty member, I chose a much less time intensive approach. At each of the nine participating sites, a key departmental administrator designated by the department chief was asked to identify all eligible care providers. The administrator was supplied written guidelines as to whom should be included (all providers from the five particular functions who were directly or indirectly involved with providing care for joint replacement patients). Surveys were mailed to all eligible care providers initially during the second month of the study period, with one repeat mailing during the study period for non-respondents. I received responses from 338 of 666 providers, for an overall response rate of 51%. The mailed survey approach resulted in an acceptable response rate, but one that was far lower than the response rate when the survey was administered in person.
Scoring the Responses and Constructing the Relational Coordination Measure
Relational coordination is first constructed for each individual respondent, seeing each respondent as the center of his or her own relational coordination network. As we will see below, if analyses support the proposition that relational coordination is significantly different across sites in your sample, you can then aggregate to a site-level measure of relational coordination. However, relational coordination is first of all an individual respondent-level measure, measuring the connections between an individual respondent and others. Please see Exhibit 12 for a sample of survey responses from a hypothetical Respondent 13 who is a member of Function 4 at Site A.
EXHIBIT 12: Survey Responses from Respondent 13 (member of Function 4, Site A)
EXHIBIT 13 provides a summary of responses for Respondent 13, including recommended variable names. Note that the first variable is the respondent's site (A), the second variable is the respondent ID (13), and the third variable is the functional identity of the respondent (4). The relational coordination variables are taken directly from the survey above, reflecting each of the seven dimensions of relational coordination measured with respect to each of the functional groups - five in this example. If there are five functional groups in the work process, as in this example, the number of relational coordination scores for each respondent will equal 7*5 or 35.
EXHIBIT 13: Summary of Responses from Respondent 13 (member of Function 4, Site A)
Second, compute a variable for each of the seven dimensions of relational coordination. The frequency of communication, for example, will be an average of the scores reported by the respondent for each of the functional groups: Freq = mean (Freqfunc1 Freqfunc2 Freqfunc3 Freqfunc4 Freqfunc5).4 See EXHIBIT 14 below for the variable names and equations used to construct these variables. You will have seven variables for each survey respondent - one for the frequency of communication, one for the timeliness of communication, one for the accuracy of communication, and so on. Relational coordination is then constructed for each individual respondent as an equally weighted index of the 35 scores, resulting in one single score for relational coordination for each respondent.
4 Relational coordination can be constructed as the average strength of ties reported by an individual respondent, or as the percent of strong ties (4 or 5 on the 5-point scale) reported by an individual respondent. The more common approach by far is the average strength of ties, so that approach is presented here.
EXHIBIT 14: New Variables Created for Respondent 13 (member of Function 4, Site A)
Third, you may also want to look at relational coordination between particular functions in order to assess relational coordination at the dyadic level. Relational coordination with Function 1, for example, will be an average of the seven different scores reported by the respondent for Function 1: RCfunc1 = mean (Freqfunc1 Timefunc1 Accufunc1 Probfunc1 Knowfunc1 Respfunc1 Goalfunc1). This will result in five new variables for each survey respondent - one for relational coordination with Function 1, another for relational coordination with Function 2, and so on. See the lower panel of EXHIBIT 11 for the equations that are used to create these new variables.
These new variables that measure relational coordination with each individual functional group (including his or her own functional group) can be placed into a matrix diagram like the ones shown earlier in EXHIBITS 8 and 9. Because our sample respondent is a member of Function 4, his or her scores are placed in the row for Function 4. See EXHIBIT15 below for an example of how this works. As we receive additional survey responses and compute scores from the responses, these scores can also be added to our matrix diagram.
EXHIBIT 15: Matrix of Relational Coordination Ties for Respondent 16
