The trust and teamwork scale was administered as part of a larger United States Agency for International Development funded project- Innovation, Research, Operations, and Planned Evaluation for Mothers and Children (I-ROPE) currently underway. The aim of this operations research is to evaluate the development and implementation of five maternity waiting homes (MWH), connected to rural community clinics in post-conflict Liberia. The five rural clinics with MWHs are matched with five demographically similar clinics without MWHs. An eleventh clinic serving the referral hospital is included in the study to capture those women referred to the next level of care. Traditionally, MWHs have been viewed as a temporary lodging located near a hospital or clinic that is available for women with high risk complications or who live a significant distance from the health center . The MWHs in this study are available for all women irrespective of complications or distance to clinic and are utilized for stays prior to delivery, short and long antenatal stays due to complications (i.e., malaria), and postpartum stays.
Prior to construction of the MWHs, community meetings were held to determine the needs of the communities. Communities agreed the MWHs would be run by a group of TTMs from the catchment area overseen by the CM. Involvement of the TTMs from the outset in the management of the MWH lends the facilities credibility among the community. The long-term goal is to ascertain whether MWHs reduce maternal mortality, increase child survival, and increase the use of skilled midwives during delivery. The study team hypothesized increased communication in the communities with MWHs would impact the trust and teamwork between the TTMs and CMs.
The instrument described in this article utilized a modified version of a pre-existing trust and teamwork scale  to evaluate the trust and teamwork among participating TTMs and CMs in the parent study. The instrument was empirically derived through formative research in Ethiopia by Dynes using in-depth interviews with frontline lay health workers including traditional birth attendants . Institutional review board approval was obtained from the University of Michigan, Health Sciences and Behavioral Sciences Review Board and cleared with the Liberian Ministry of Health and Social Welfare.
The Trust and Teambuilding Scale was originally developed using semi-structured interviews with 30 healthcare workers in a low resource region of Ethiopia . The concepts of trust and teamwork were examined through participants’ understanding of the attributes, consequences, and conceptualization of the constructs. The original 40–item survey was tested using a four-point Likert scale with “1=Strongly disagree, 2=Disagree, 3=Agree, and 4=Strongly agree” using cards with symbols. Due to the low literacy of the population, participants pointed to a circle indicating “agree” or a square indicating “disagree.” Next, the interviewer opened the folded card to reveal a big circle and a little circle if the participant agreed or a big square and little square if the participant disagreed. The respondent was then questioned as to whether they agreed or disagreed a lot or a little. Use of these response sets with the Trust and Teambuilding Scale demonstrated strong internal consistency (Cronbach alpha’s of .79 to .81) and cultural adequacy to assess trust and teamwork in a population of 197 community level health workers in Ethiopia (M. Dynes, Personal Communication, January 30, 2012).
Sample and setting
Data for this exploratory descriptive study were obtained from a sample of TTMs (n=90) in the 11 rural communities participating in the I-ROPE parent study. The setting for this study is one rural county in north central Liberia with a total population of over 300,000. The population within the catchment communities of the study is 109,000 with 25,000 women of reproductive age.
The official language of Liberia is English. The literacy rate in the adult population is approximately 44% . Inclusion criteria included TTMs providing support to pregnant women in the catchment area of the clinics, participating in the parent study, age 18 or older, and able to speak English or one of the native tribal languages, Kpelle or Mano. All available TTMs participating in the parent study were invited to participate in the study. There were no declinations.
The study took place between March 2012 and April 2012. The original survey was written in English but included Ethiopian phrases to describe some of the concepts. Prior to survey administration in Liberia, the study team first removed the descriptive Ethiopian phrases from the original survey. When these descriptive Ethiopian phrases were removed, two questions became very similar with both describing the same concept of honesty. To ensure clarity and avoid repeating the same concept, one of these questions was discarded prior to using the survey in Liberia resulting in a total of 39 items.
Two female Liberian study team members, the research nurse for the I-ROPE study and the reproductive health supervisor for the County Health Team, were present for the administration of every survey. They served as the cultural brokers and facilitated translation from English to Kpelle. In addition, in one catchment community, four participants identified Mano as their primary language. At this location a member of the clinic staff (who was not a CM) assisted in translation alongside our approved translators.
Prior to the beginning of data collection, the purpose of the study – to understand trust and teamwork between TTMs and CMs – was explained to participants. Confidentiality was assured and verbal informed consent obtained. Participants were informed they could decline to answer any question and their answers would only be shared in aggregate.
Initially, our team conducted the survey in its original format utilizing individual interviews with TTMs that included one Liberian translator and a female study team member to record the answers that the respondent indicated using the four-point Likert scale with picture cards. However, after piloting this procedure at the first community (n=11), it was determined this process was very lengthy and burdensome to a low-literate population. Additionally, participants had difficulty grasping the meaning of the four-point Likert scale. We assessed their ability to understand ‘agree and disagree’ or ‘yes and no’ using the cards and found dichotomous scoring much easier for comprehension.
In the next community, the research team conducted the survey in a group setting using dichotomous scoring of agree or disagree. Each participant was paired with a female study team member working to record the answers in a private space so others could not see her answers. The questions were read by one of the Liberian translators. Following the question, the participant silently pointed to her response, a circle for agree and a square for disagree, which was then recorded by the research team member.
We also noted the questions written in a ‘double-negative’ format were extremely confusing for the participants. We continued to use the full 39-item Trust and Teambuilding Scale in three more communities with a total of 48 participants. The administration of the survey with paired research team members decreased the time commitment of participants; taking approximately 45–60 minutes to complete the full survey. At this point in the study, the research team decided to take a closer look at the questions to determine if the survey could be utilized in an abbreviated version. We decided to make three major changes: 1) eliminate the questions with little to no variability; 2) eliminate any questions written in a double-negative format; and 3) use wording that allowed participants to respond to questions related to their own personal relationship as well as the relationship of other TTMs with CMs in their community. This abbreviated dichotomous version of the instrument, or the Trust and Teamwork Scale – Liberia, included eight items to assess trust and ten items to assess teamwork modified from the previous scale.
We were able to use the abbreviated Trust and Teamwork Scale – Liberia in six communities with an additional 42 participants. During the administration of the abbreviated version we found the length of time to administer the survey decreased dramatically to approximately 20–30 minutes, participants remained engaged throughout the entire survey, and the dichotomous version was easier for participants to understand. During this process we also noted how the participants responded to the survey. For instance, we determined the translator and participants were describing the responses as ‘truth or lies’ instead of ‘yes or no’ or ‘agree or disagree.’ The participants also found it easier to equate the circle and square on the picture cards to items they were more familiar with; describing the circle as a ‘ball’ and the square as a ‘basket.’
Procedures for data analysis
For the data analysis, variables were all coded with positive numbers representing more trust and teamwork. New testing of psychometric properties is warranted when a non-standardized measure is changed for subsequent use and is essentially altered into a different measure . This section will report on the reliability and validity of the modified Trust and Teamwork Scale for use in Liberia.
Exploratory factor analysis was applied to consolidate items and identify the factors within the Trust and Teamwork Scale - Liberia. Due to the dichotomous nature of the observed variables, it was decided to conduct the analysis in Mplus software (Mplus version 6.1) . Mplus is useful because it applies a probit in the place of ordinary least squares, which is important in the analysis of dichotomous variables as a probit does not depend on normal distribution. Because Mplus uses a probit regression of the item on the factor, it allows for analysis of non-linear relationships .
Prior to beginning factor analysis the suitability of the sample size was examined. The full survey elicited a total sample size of 42, with an additional 48 completing a portion of the questions in the Trust and Teamwork Scale – Liberia. Several questions did not display any variability among participants. Since Mplus is not able to analyze questions without variability, it was determined a priori to exclude these questions from data analysis. This resulted in 16 items for our factor analysis.
Due to our small sample size it was determined missing data would be handled with the default setting in Mplus using all available data to estimate the model . Only one item had missing data, the question “In your village, the CM(s) push their work onto others” was skipped or omitted by one participant in the sample. The literature reports a range of necessary cases per item required for adequate factor analysis ranging from 4 to 10 cases per measure [29–31]. More recently, MacCallum and colleagues  took a unique approach to the problem and used factor analytic theory  to show it may not be possible to derive a minimum sample size that is appropriate in all situations. Using this decision-making, Guadagnoli and Velicer  determined that if components have four or more variables with loadings above .60, then the factor structure or pattern may be interpretable despite the sample size. Exploratory factor analysis with small samples (≤50) have been demonstrated to be a feasible undertaking  especially when the data exhibit high loadings, a low number of factors, and high numbers of variables per factor [36, 37]. Regardless of the sample size regulations employed, the sample size for this factor analysis was small and limited by the number of available TTMs in our setting. Therefore, a second exploratory factor analysis was employed with a portion of the variables to confirm the initial results.
The construct validity of the Trust and Teamwork Scale – Liberia was examined using weighted least squares with mean variance and varimax rotation. In addition to using Mplus for statistical analyses, the Statistical Package for the Social Sciences (SPSS) version 19.0 was utilized for descriptive statistics, reliability, and validity testing. Additional correlational analyses were completed to ascertain the presence of contrast and convergent validity of the revised measure.