Lohmann et al. Human Resources for Health (2017) 15:33 DOI 10.1186/s12960-017-0208-1 METHODOLOGY Open Access Measuring health workers’ motivation composition: validation of a scale based on Self-Determination Theory in Burkina Faso Julia Lohmann1*, Aurélia Souares1, Justin Tiendrebéogo2, Nathalie Houlfort3, Paul Jacob Robyn4, Serge M. A. Somda5 and Manuela De Allegri1 Abstract Background: Although motivation of health workers in low- and middle-income countries (LMICs) has become a topic of increasing interest by policy makers and researchers in recent years, many aspects are not well understood to date. This is partly due to a lack of appropriate measurement instruments. This article presents evidence on the construct validity of a psychometric scale developed to measure motivation composition, i.e., the extent to which motivation of different origin within and outside of a person contributes to their overall work motivation. It is theoretically grounded in Self-Determination Theory (SDT). Methods: We conducted a cross-sectional survey of 1142 nurses in 522 government health facilities in 24 districts of Burkina Faso. We assessed the scale’s validity in a confirmatory factor analysis framework, investigating whether the scale measures what it was intended to measure (content, structural, and convergent/discriminant validity) and whether it does so equally well across health worker subgroups (measurement invariance). Results: Our results show that the scale measures a slightly modified version of the SDT continuum of motivation well. Measurements were overall comparable between subgroups, but results indicate that caution is warranted if a comparison of motivation scores between groups is the focus of analysis. Conclusions: The scale is a valuable addition to the repository of measurement tools for health worker motivation in LMICs. We expect it to prove useful in the quest for a more comprehensive understanding of motivation as well as of the effects and potential side effects of interventions intended to enhance motivation. Abstract French Contexte: La motivation des agents de santé dans les pays à revenu faible et intermédiaire est devenue un sujet de grand intérêt pour les décideurs et les chercheurs au cours des dernières années. Pourtant, de nombreux aspects de la motivation des agents de santé ne sont pas encore bien compris. Ceci est dû en particulier à l’absence d’outils de mesure appropriés. Cet article présente une preuve de la validité conceptuelle d’une échelle psychométrique développée pour mesurer la composition de la motivation, c’est-à-dire le degré auquel des types de motivation d’origine différente à l’intérieur et à l’extérieur d’une personne contribuent à leur motivation globale au travail. L’échelle est fondée sur la théorie de l’auto-détermination (Self-Determination Theory). (Continued on next page) * Correspondence: julia.lohmann@uni-heidelberg.de 1 Institute of Public Health, Faculty of Medicine, Heidelberg University, Im Neuenheimer Feld 324, Heidelberg, Germany Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Lohmann et al. Human Resources for Health (2017) 15:33 Page 2 of 12 (Continued from previous page) Méthodes: Une enquête transversale a été mise en place auprès de 1 142 infirmières dans 522 formations sanitaires gouvernementales de 24 districts du Burkina Faso. Par analyse factorielle confirmatoire, nous avons examiné si l’échelle mesure ce qu’elle était supposée mesurer (validité structurelle et convergente/discriminante) et si ses propriétés de mesures sont comparables dans différentes sous-groupes d’agents de santé (invariance de la mesure). Résultats: Les résultats montrent que l’échelle mesure une version légèrement modifiée du continuum de motivation proposée par la théorie de l’auto-détermination. Les propriétés de mesure étaient globalement comparables entre les sous-groupes, mais une certaine prudence est indiquée si une comparaison des moyennes entre les groupes est l’objectif principal de l’analyse. Conclusion: L’échelle est une un apport important au référentiel des outils de mesure de la motivation des agents de santé dans les pays à revenu faible et modérés. Elle sera utile pour une meilleure compréhension de la motivation des prestataires, ainsi que des effets positifs et potentiellement secondaires des interventions visant à renforcer la motivation. Keywords: Health worker motivation, Motivation composition, Measurement, Validation, Self-Determination Theory Background Determination Theory (SDT) [4, 14] and was developed Recent years have witnessed an increased awareness of the for use in questionnaires or structured interviews. It paramount importance of a motivated health workforce assesses general motivation towards work rather than for the functioning of health systems, particularly in coun- task- or situation-specific motivation. The article presents tries burdened by severe resource limitations [1]. Inter- evidence for the scale’s validity from a structured survey ventions targeting health worker motivation such as with nurses in Burkina Faso. Table 1 contains our specific performance-based financing (PBF) have become ex- research questions. tremely popular among policy makers in low- and middle- income countries (LMICs) [2, 3]. Despite the attention The self-determination continuum of motivation such interventions are receiving, gaps in understanding re- Self-Determination Theory was introduced in the mid- main. In particular, the mechanisms through which inter- 1980s as a general framework of human motivation [14] ventions bring about motivational changes and potential and has since been extensively studied and further refined side effects thereof remain poorly understood [4–8]. For [15]. As part of the overall theory, SDT proposes the self- instance, there is an ongoing debate around whether the determination continuum of motivation (Fig. 1), a tax- monetary incentives involved in PBF undermine intrinsic onomy of five major dimensions of motivation that are motivation (“crowding out effect”) [5]. distinguished by the extent to which they stem from con- The limited availability of context-adapted research tingencies outside the person (controlled motivation) or tools to study motivation is a major factor contributing to originate within the person (autonomous motivation) [16]. this knowledge gap. Research on health worker motivation in LMICs has mostly focused on the overall amount or on determinants and outcomes of motivation, leaving other Table 1 Aspects of validity investigated and specific research relevant dimensions discussed in the psychological litera- questions ture such as motivation composition relatively unexplored Type of validity Research questions [4, 9]. Corresponding quantitative measurement tools Structural validity RQ1: Is the assumed internal theoretical structure (e.g., [10–13]), while without doubt useful to answer many of motivation (i.e., the SDT continuum of motivation; research questions, are not suited to others, including that Fig. 1) represented in the around the crowding out effect which deals with a shift in data as it was intended during scale development? a. Do respondents distinguish the five dimensions of motivation composition from intrinsic to extrinsic forms. motivation? Against this background, this article contributes to b. Are adjacent dimensions more closely related than expanding the methodological repository for health non-adjacent dimensions? worker motivation research by presenting evidence on Generalizability RQ2: Do psychometric properties and interpretations the construct validity of a newly developed psychometric generalize across health worker subgroups (measurement invariance)? scale to measure health worker motivation composition. Convergent and RQ3: To what extent do relationships between the We define motivation composition as the extent to discriminant motivation measure and measures of other related which motivation of different origin within and outside a validity constructs correspond to what is theoretically person contributes to their overall work motivation. The expected and has been found in previous research with other, established measures? scale is theoretically grounded in Deci and Ryan’s Self- Lohmann et al. Human Resources for Health (2017) 15:33 Page 3 of 12 Fig. 1 The self-determination continuum of motivation. Legend: adapted from [15, 16] The scale validated in this article measures these five (e.g., [4, 7, 8, 10–13, 19–25]). For a theoretical applica- motivation dimensions. Motivation originating fully tion of SDT and the taxonomy to LMIC healthcare set- within the person, such as pure enjoyment of a task, is tings, see [4]. termed intrinsic motivation in SDT. Extrinsic motiv- ation, in contrast, refers to motivation derived from an Methods instrumental purpose of behavior. External regulation Study context corresponds to what is usually referred to as extrinsic Burkina Faso’s healthcare delivery system relies primarily motivation: the wish to attain or avoid some conse- on the public sector which manages approximately 80% of quence. SDT differentiates three additional dimensions healthcare facilities. Primary healthcare services are mostly of extrinsic motivation by the degree to which the asso- provided by nurses, midwives, and assistant nurses and ciated contingencies have become part of the person’s midwives. Like many other LMICs, Burkina Faso’s health self: introjected regulation refers to motivation derived system suffers from multiple challenges including a short- from self-pride, reputation, or feelings of duty, identified age of certain health worker cadres, their unequal geo- regulation to motivation driven by recognition of the im- graphical distribution, and challenging working conditions portance of one’s job, and integrated regulation to full including low pay, substandard infrastructure and equip- congruency between one’s personal goals and values and ment, poor supervision, shortages in drugs and other sup- those of one’s job. They differ from external regulation plies, and few incentives for individual high performance in that they do not need to be maintained from the out- [26–28]. In 2014, the Ministry of Health with support side through rewards or punishment. However, they are from the World Bank implemented a PBF pilot interven- not fully intrinsic as corresponding behavior is instru- tion to strengthen the healthcare system by addressing mental in catering to a person’s set of values and goals some of these challenges. Our study took place in the con- rather than performed out of pure interest or enjoyment. text of the impact evaluation of this intervention. A large body of research has linked autonomous forms of motivation to more favorable performance and other Motivation composition measure outcomes (e.g., wellbeing, organizational commitment) The psychometric scale to measure motivation composition than controlled forms of motivation [9, 15, 17]. was developed by our research team prior to the validation The validity and usefulness of the SDT taxonomy has study presented in this article. A detailed description of this been confirmed in a wide range of work settings, process can be found in Additional file 1. A pretest con- although mostly in North America and Europe [15]. firmed the scale’s content validity, supporting the validity of However, the few studies from LMIC (non-healthcare) the SDT taxonomy in the context and affirming that the settings [18] and the (non-SDT-based) literature on items cover the five motivation constructs well and in health worker motivation in LMICs suggest its validity context-appropriate language. in LMIC healthcare contexts as well. Specifically, sources Similar to other SDT-based measures (e.g., [18, 29]), of motivation identified by the latter correspond well to the scale’s measurement rationale is grounded in the idea the five dimensions differentiated by the SDT taxonomy that individuals will reveal their underlying motivation Lohmann et al. Human Resources for Health (2017) 15:33 Page 4 of 12 composition in the reasons for the actions they provide. interview approximately 80%, resulting in a total sam- Following an introduction, a reflective exercise, and a ple size of 1142 (per facility: mean = 2.2, sd = 1.6, min = guiding question (“Why are you motivated to work?”), 1, max = 11). In addition to the motivation scale, the respondents are thus presented with 26 reasons for which survey contained questions on training, clinical know- they might be motivated to work (4–12 per motivation ledge, compensation, and working conditions. Data was dimension; see Additional file 1). They are asked to indi- collected on paper and digitalized using a double data cate, on an 11-point scale and with a visual aid, the degree entry strategy. Table 2 shows the sample distribution to which each of these reasons are important for their per- on key characteristics. sonal work motivation. Respondents’ answers are then used to derive an estimate of their underlying motivation Structural validity analyses level on the five dimensions. The structural validity analyses (research question 1 In order to counteract diverse response biases, we (RQ1)) aimed to confirm that the scale measures the mo- used a hybrid mode of administration, with interviewers tivation dimensions of the SDT continuum as intended. reading out instructions and items but interviewees re- We first conducted a thorough integrated semantic and cording their own answers on a separate questionnaire psychometric item analysis (including inspection of item copy. The questionnaire was administered in French in distribution and correlation patterns; in Stata 12), in re- light of the high French proficiency level of Burkinabé sponse to which we excluded 8 items from the initial 26- health workers. One explicit aim of the validation ana- item scale due to suboptimal psychometric properties or lyses was the selection of a subsample of items for a final phrasing (see Additional file 2). The remaining 18 items shorter and easy-to-administer scale. were subsequently subjected to a confirmatory factor ana- lysis using structural equation modeling (SEM). In line Sample with standard SEM terminology, we refer to the five mo- We assessed the scale’s validity with data from a structured tivation dimensions as “factors” from here forward. We health worker survey implemented between October 2013 tested the five-factor model corresponding to Fig. 1 and March 2014 in the context of the abovementioned against the three theoretically viable alternative models in PBF impact evaluation baseline. The sampling strategy was Table 3, which emerged as alternative taxonomies during aligned with the cluster sampling strategy of the impact the scale development process or have shown good evaluation accordingly [30]. Data was collected from ap- model-data fit in previous research (e.g., [18, 31]). All proximately two thirds of all government health facilities modeling was performed with Mplus 7.31, using a max- in 24 districts of six regions of the country. Research assis- imum likelihood estimator with robust standard errors to tants were instructed to interview all nurses, midwives, account for our non-normal data distribution. Standard and assistant nurses and midwives in 498 primary as well errors were adjusted according to the clustered sample as selected staff in 24 secondary-level facilities present on structure. Missings were handled with Mplus’ standard full the day of the study team visit. Fifty-five per cent of all information procedure. All factors were allowed to covary. nursing and midwifery staff were on duty and present on No cross-loadings or correlated item residuals were speci- the day of facility visit. Of those, interviewers were able to fied to facilitate interpretation in light of potential use of Table 2 Sample characteristics Variable Number Per cent Mean SD Median Min Max Sex Female 646 56.6 Male 496 43.4 Age 34.4 5.4 33.5 20 56 Seniority (years in healthcare service) 6.2 5.0 4.5 0 36 <5 years 504 44.1 ≥5 years 638 55.9 Health worker type Nurse/Midwife (diploma) 495 43.4 Nurse/Midwife (assistant) 647 56.6 Total 1142 100.0 Legend: “Nurse/Midwife (diploma)” includes the following cadres: Attaché de santé (specialist nurse), Infirmier Diplômé d’Etat (nurse with state diploma), Sage- Femme/Maïeuticien d’Etat (midwife with state diploma), and Infirmier Breveté (licensed nurse); “Nurse/Midwife (assistant)” includes the following cadres: Accoucheuse Auxilliare (assistant midwife) and Accoucheuse Brevetée (licensed midwife) Lohmann et al. Human Resources for Health (2017) 15:33 Page 5 of 12 Table 3 Alternative models tested α. We eliminated 3 further items in the model-fitting Model A Five-factor model corresponding to Fig. 1 process (see Additional file 2), arriving at the final 15-item Model B Four-factor model, combining the integrated scale in Table 4. All results presented in this article are and identified types of regulation which have based on this final 15-item scale. proven difficult to separate in previous research Model C Five-factor model as Model B but dividing external Generalizability analyses regulation into a social and an economic subfactor The generalizability analyses aimed to confirm that the Model D Two-factor model, differentiating autonomous scale measures the same motivation dimensions equally (intrinsic motivation, integrated/identified regulation; AUT) and controlled (introjected, external regulation; well in different sample subgroups (“measurement in- CTRL) motivation variance”; RQ2). This is a necessary requirement for later substantive analyses aiming to compare motivation across different health worker subgroups. Specifically, the scale with composite scores. Models were evaluated we tested the scale for invariance across sexes, seniority with standard fit indices, including χ2, comparative fit levels, and qualification levels. Following the steps out- index (CFI), standardized root mean square residual lined in Table 5 [33], Model C was simultaneously esti- (SRMR), and root mean square error of approximation mated in each respective subgroup, with an increasing (RMSEA), and compared to each other with the Akaike number of parameters restricted to equality between information criterion (AIC) [32]. For the best-fitting subgroups in each testing step. The scale is a measure- model C, we inspected all model parameters, Mplus’ ment invariant at each level when the added equality re- modification indices, factor correlations, and Cronbach’s strictions do not lead to significantly worse model fit Table 4 Final item list and descriptive statistics Item Number Mean sd p50 Max Min Intrinsic im1 Parce que j’aime faire ce que je fais chaque jour au travail. 1 139 7.91 2.49 9 10 0 motivation (IM) Because I enjoy doing what I do at work every day. im2 Parce que mes tâches au travail me plaisent beaucoup. 1 142 8.21 2.09 9 10 0 Because I enjoy my work tasks. im3 Parce que le travail que je fais est très intéressant. 1 139 8.22 2.09 9 10 0 Because the work that I do is very interesting. Integrated/identified iden1 Parce qu’être un agent de santé est un élément fondamental de ce que je suis. 1 134 8.08 2.29 9 10 0 regulation (IDEN) Because being a health worker is a fundamental part of who I am. iden2 Parce que mon travail est extrêmement important pour mes patients. 1 137 8.53 1.89 9 10 0 Because my work is extremely important for my patients. iden3 Parce que je veux changer quelque chose dans la vie des autres. 1 138 7.90 2.55 9 10 0 Because I want to make a difference in people’s lives. Introjected intro1 Pour avoir une bonne opinion de moi-même. 1 141 7.45 2.70 8 10 0 regulation (INTRO) In order to feel good about myself. intro2 Parce que ma réputation dépend de mon travail. 1 133 7.19 3.00 8 10 0 Because my reputation depends on my work. External ext1 A cause de la reconnaissance que je reçois de mes patients et de la communauté. 1 132 6.32 3.21 7 10 0 regulation-social Because of the appreciation I receive from my patients and the community. (EXT-S) ext2 Pour ne pas laisser tomber mon équipe. 1 136 4.86 3.18 5 10 0 So I do not let my team down. ext3 Parce que mon responsable direct reconnaît mon travail et m’apprécie. 1 128 6.22 3.17 7 10 0 Because my supervisor recognizes and appreciates me. External ext4 A cause des avantages liés à mon travail. 1 137 3.75 3.29 4 10 0 regulation-economic Because of the benefits that come with my job. (EXT-E) ext5 Pour pouvoir subvenir aux besoins de ma famille. 1 141 6.50 3.03 7 10 0 In order to be able to provide for my family. ext6 Parce que mon travail me procure la sécurité financière. 1 136 4.76 3.10 5 10 0 Because of the financial security my job provides me with. ext7 Afin de gagner de l’argent. 1 134 3.67 3.17 3 10 0 In order to earn money. Legend: The English translation is intended to facilitate understanding for the non-French-speaking readership. It is not tested and validated and might thus not be perfectly equivalent to the French version Lohmann et al. Human Resources for Health (2017) 15:33 Page 6 of 12 Table 5 Measurement invariance testing steps [33] Test for Interpretation Model constraints Configural invariance Tests for the assumption of the same underlying No specific constraints are imposed on the factor structure in all subgroups, i.e., the overall estimated parameters. model fits the data similarly well in all subgroups Metric invariance Tests whether the same constructs are measured • Factor loadings estimated freely but constrained across subgroups, i.e., whether respondents in to equality in the subgroups different subgroups attribute the same meaning to the respective motivation factors Scalar invariance Tests whether subgroups can be compared on • Factor loadings estimated freely but constrained their mean scores or whether subgroups score to equality in the subgroups systematically different (at the same underlying • Item intercepts estimated freely but constrained to level of motivation) for certain items equality in the subgroups Residual variance invariance Tests whether the proportion of contamination by • Factor loadings estimated freely but constrained to other constructs as measured by the different items equality in the subgroups (i.e., variance that is not explained by the intended • Item intercepts estimated freely but constrained to factors) is equal across groups and whether equality in the subgroups measurements are thus fully comparable across groups • Item residual variances estimated freely but constrained to equality in subgroups compared to the respective less restricted model. Nested Details on hypotheses and measurement of the external model comparisons were conducted with the rescaled constructs are provided in Table 6. We built a separate likelihood ratio test [34]. model for each external construct by adding a measure- ment model for the respective construct to Model C, Convergent/discriminant validity analyses allowing the external construct factor to covary with all The convergent/discriminant validity analyses aimed to five motivation factors. provide further evidence that the scale measures the SDT taxonomy as intended by relating motivation with Results constructs for which relationships with the SDT motiv- Structural validity ation dimensions are relatively well established (RQ3). If The structural validity analyses aimed to confirm that the the new scale does indeed measure what it is intended to scale does indeed measure the different motivation dimen- measure, relationships with external constructs should sions of the SDT continuum. We intended to test the approximately correspond to those found in previous “pure” SDT model (Fig. 1; Model A in Table 3) against research, contextual differences taken into consideration. three theoretically viable alternative models. Unfortunately, Specifically, we related motivation to organizational sup- Model A could not be estimated with the final subset of port, organizational commitment, and intentions to quit. items as we were only able to retain one integrated Table 6 Convergent/discriminant validation constructs and hypotheses (based on SDT and previous research [15, 18, 43, 44]) Construct and hypotheses Measurement Organizational support: extent to which respondents feel supported by Organizational support was measured with six items partly adapted their supervisor and coworkers, both technically and emotionally. from [45, 46] (Cronbach’s α = .90) Hypotheses: Item examples: “The people I work with are there to help me when (a) Autonomous (intrinsic) types of motivation are closely and positively I need support.”; “I can absolutely rely on the people I work with.” related to organizational support. Response scale: 0 (do not agree at all)–10 (completely agree) with (b) Controlled (extrinsic) types of motivation are unrelated to organizational visual aid (analogous to the motivation measure) support. Organizational commitment: extent to which respondent are emotionally Organizational commitment was measured with three items partly attached to their workplace adopted from [13, 47] (α = .74) Hypotheses: Item examples: “I would not want to work for a different health facility. (a) Autonomous types of motivation are closely and positively related to ”; “I am proud to be working for this health facility.” organizational commitment. Response scale: 0 (do not agree at all)–10 (completely agree) with visual (b) Controlled types of motivation are unrelated to organizational aid (analogous to the motivation measure) commitment. Intentions to quit: extent to which respondents would like to leave their Intentions to quit were measured with three items partly adopted from current position [11] (α = .72) Hypotheses: Item examples: “I often feel like leaving my job.”; “Accepting to work for (a) Autonomous types of motivation are negatively related to turnover this facility was a mistake.” intentions. Response scale: 0 (do not agree at all)–10 (completely agree) with visual (b) Controlled types of motivation are positively related to turnover aid (analogous to the motivation measure) intentions. Lohmann et al. Human Resources for Health (2017) 15:33 Page 7 of 12 regulation item. Table 7 presents fit statistics for the three seniority, and health worker qualification level. The scale alternative models. Model C, which combines the inte- is fully invariant for seniority in healthcare. Only partial grated and identified dimensions but differentiates external measurement invariance could be established for sex. regulation into a social and an economic subcomponent, Specifically, women scored higher than men on intro1 clearly demonstrated the best fit. χ2 was significant as ex- and ext6, but lower on intro2, at the same underlying pected given our relatively large sample size, high factor levels of introjected and external regulation, respectively correlations, and non-normally distributed data [32, 35] but (scalar non-invariance). This raises concerns about of a magnitude that does not warrant concerns for model factor means comparability for the concerned subscales. fit. All other fit indices were good in absolute terms, indi- However, as intro1 and intro2 are biased in opposite cating that the modified five-factor model is well repre- directions in around the same magnitude, we can as- sented in the data. All following results thus pertain to sume biases to cancel each other out. For ext6, consider- Model C. A graphic representation including standardized ing that it is only one of four items measuring economic coefficients for all estimated parameters as well as modifica- external regulation and the systematic difference in scor- tion indices is given in Additional file 2. For each motiv- ing is relatively small, we can also assume that the over- ation factor, item-factor loadings are of relatively similar all bias is of little practical relevance [33]. We could also magnitude; the items thus indicate the respective factor establish only partial scalar invariance for qualification with similar strength. Modification indices signal that some level. Item ext7 had a somewhat higher factor loading items, particularly ext6 and ext7, load on factors other than (i.e., item is more strongly indicative of factor) in fully the intended to some extent. Overall, however, such cross- qualified than in assistant nurses (metric non- loadings are low in magnitude, indicating good item dis- invariance). At the scalar level, fully qualified nurses sys- criminatory power. Although also mostly low in magnitude, tematically scored higher on intro1, ext7, and im3 and modification indices show many residual (error term) lower on intro2. In similar lines of reasoning as for sex, correlations, particularly for the external regulation we can reasonably assume that these systematic differ- (EXT) items. Factor correlations (Table 8) display the ences do not majorly threaten comparability between expected simplex pattern, i.e., decreasing magnitude groups substantially, however. with decreasing conceptual closeness. Cronbach’s α is relatively low for all factors. Convergent/discriminant validity The convergent/discriminant validity analyses aimed to Generalizability provide additional evidence that the scale measures what The measurement invariance analyses aimed to confirm it was intended to measure by relating motivation to that the scale has the same measurement properties in other variables with which the relationship is well estab- different subsamples and that measurements (scores, lished. Table 10 shows correlations of the motivation variances, etc.) can thus be compared between health factors with the three constructs introduced in Table 6. worker subgroups. Table 9 shows the results for sex, Correlation patterns are generally in the expected Table 7 Results of the structural validation analyses Model χ2 df p RMSEA p RMSEA ≤.05 CFI SRMR AIC A Model A, the original five-factor model corresponding to Fig. 1, could not be estimated as only one integrated regulation item was retained in the fitting process (at least two are necessary for model identification) B Four-factor model: 472 84 .000 .064 .000 .867 .069 78 649 IM (im1-im3), IDEN (iden1-iden3), INTRO (intro1 intro2), EXT (ext1-ext7) C Five-factor model: 227 80 .000 .040 .996 .950 .033 78 318 IM (im1-im3), IDEN (iden1-iden3), INTRO (intro1 intro2), EXT-S (ext1-ext3), EXT-E (ext4-ext7) D Two-factor model: 677 89 .000 .076 .000 .799 .076 78 927 AUT (im1-iden3), CTRL (intro1-ext7) Interpretation of fit indices [32]: Insignificant χ2 values indicate good model-data fit. However, due to a number of conceptual and statistical issues, χ2 is often significant even in the case of a relatively good model fit. CFI values approaching .95 as well as RMSEA values of .05 or smaller and SRMR values of .05 and smaller are considered indicative of good model fit. Smaller AIC values indicate better data-model fit compared to alternative models (evaluation goodness of fit (likelihood function) versus complexity of the model) Legend: IM intrinsic motivation factor, IDEN integrated/identified regulation factor, INTRO introjected regulation factor, EXT external regulation factor, EXT-S external regulation-social factor, EXT-E external regulation-economic factor, AUT autonomous motivation factor, CTRL controlled motivation factor Lohmann et al. Human Resources for Health (2017) 15:33 Page 8 of 12 Table 8 Model-estimated factor correlation matrix and reflecting realities in the specific context rather than Cronbach’s α (on the shaded diagonal cells) for the motivation being indicative of measurement issues, however [6]. factors in Model C IM IDEN INTRO EXT-S EXT-E Discussion The paper presents evidence on the validity of a newly IM .64 developed scale to measure motivation composition of IDEN .87 .66 health workers, i.e., the relative contribution of different INTRO .72 .82 .58 kinds of motivation to their overall work motivation, from a sample of nurses in Burkina Faso. EXT-S .60 .64 .86 .58 Our findings show that the scale measures a somewhat EXT-E .25 .23 .51 .62 .75 modified version of the SDT continuum of motivation well and relatively consistent in different health worker sub- All correlation coefficients are Person correlations and significantly different groups. Specifically, our analyses suggest that the scale is from zero Legend: IM intrinsic motivation, IDEN integrated/identified regulation, INTRO not able to distinguish between integrated and identified introjected regulation, EXT-S external regulation-social, EXT-E external regulation- regulation. This finding is in line with what emerged during economic the scale development process and with previous attempts to measure the SDT continuum [18, 29]. From an applied directions, supporting the notion that the scale measures perspective, not distinguishing the two dimensions is even the SDT continuum of motivation well. Organizational advantageous insofar as policy implications are similar and support and organizational commitment are more interpretation thus facilitated. Our analyses further suggest strongly related to introjected regulation than expected to separate external regulation into a social dimension, in- based on previous research. Correlations of all motiv- cluding aspects of social interaction and recognition, and ation factors with intentions to quit are weaker than an economic dimension, pertaining to the economic secur- expected. These findings are likely substantive findings ity one’s job provides. Again, such a distinction is sensible Table 9 Measurement invariance testing results Absolute model fit Likelihood ratio test info and results χ 2 df p RMSEA p RMSEA ≤.05 CFI SRMR #free parms LR (with model df p (.05) above) Baseline model C 227 80 .000 .040 .996 .950 .033 – – – – Sex Configural invariance 333 160 .000 .044 .994 .943 .041 110 – – – Metric invariance 344 170 .000 .042 .975 .943 .043 100 9.40 10 0.50 Scalar invariance 386 180 .000 .045 .917 .932 .045 90 50.44 10 < 0.01 Scalar invariance, partial 356 178 .000 .042 .983 .941 .044 92 11.19 8 0.19 Residual variance invariance 369 191 .000 .040 .995 .941 .048 79 16.30 13 0.23 Seniority Configural invariance 332 160 .000 .043 .950 .943 .039 110 – – – Metric invariance 342 170 .000 .042 .979 .943 .042 100 9.23 10 0.51 Scalar invariance 350 180 .000 .041 .993 .944 .043 90 4.29 10 0.93 Residual variance invariance 366 195 .000 .039 .998 .944 .052 75 19.39 15 0.20 Qualification level Configural invariance 319 160 .000 .042 .980 .947 .039 110 – – – Metric invariance 338 170 .000 .042 .984 .945 .044 100 18.43 10 0.05 Scalar invariance 371 180 .000 .043 .966 .937 .046 90 37.07 10 < 0.01 Scalar invariance, partial 349 177 .000 .041 .989 .943 .045 93 9.26 7 0.25 Residual variance invariance 363 192 .000 .040 .998 .944 .048 78 18.91 15 0.22 Legend: Interpretation of the absolute model fit indices [32]: Insignificant χ2 values indicate good model-data fit. However, due to a number of conceptual and statistical issues, χ2 is often significant even in the case of relatively good model fit. CFI values approaching .95 as well as RMSEA values of .05 or smaller and SRMR values of .05 and smaller are considered indicative of good model fit Interpretation of the likelihood ratio test statistics: #free parms is the number of freely estimated model parameters; these are gradually restricted in the invariance testing process as parameters are forced to equality in the compared subgroups (see Table 5). LR (with above model and its degrees of freedom) is the χ2-distributed test statistic of the rescaled likelihood ratio test. In each row, it refers to the difference in fit of the respective model and the next less restrained (i.e., above) model. Statistical insignificance indicates that the more restricted model fits similarly as the above less restricted model, i.e., that the added parameter equality restrictions for the compared sample subgroups do not substantially worsen model fit and that the scale can thus be considered measurement invariant for the compared groups at the respective level Lohmann et al. Human Resources for Health (2017) 15:33 Page 9 of 12 Table 10 Convergent/discriminant validation results: model- which, although mostly small, likely also contributed to in- estimated factor correlations of motivation dimensions with flated factor correlations. They might have partially been external constructs caused by the more specific item phrasing compared to IM IDEN INTRO EXT-S EXT-E other SDT-based measures [18, 29]. Cross-loadings and Organizational support .46 .43 .37 .47 .12 residual correlations are often explicitly modeled to im- Organizational commitment .58 .54 .37 .38 .05a prove overall model fit, for instance, in exploratory struc- Intentions to quit −.15 −.07 a a .06 a .03 .18 tural equation models (ESEM) [32, 40]. In light of our a already good fit, we opted against doing so based on the Legend: not statistically significantly different from zero assumption that future users of the scale might want to from an applied point of view in light of the different policy analyze data using composite scores, which would be diffi- implications related to the two dimensions. The modified cult with a scale “calibrated” in an ESEM framework. taxonomy measured by the scale is visualized in Fig. 2. Limitations Methodological discussion Measurement reliability and sensitivity Our results are generally as expected. The structural and We were unable to examine measurement reliability (i.e., convergent/discriminant validity analyses support that accuracy and consistency) in-depth within the scope of the scale measures the SDT taxonomy of motivation, al- our study, beyond what was possible in the scale develop- beit in slightly modified form as explained above. It does ment process. We thus cannot exclude that respondents’ so equally well for different health worker subgroups, scores are to some extent influenced by random or sys- although with some caveats (see below), indicating that tematic measurement error rather than solely by under- the scale can be used for between-group comparisons. lying levels of motivation. The convergent/discriminant However, two aspects deserve further discussion. First, validity analysis results, specifically their consistency with despite good overall fit of the data to the five-factor model, previous research, imply that random measurement error we found relatively low levels of Cronbach’s α for all fac- is at acceptable levels. Based on the continued high scores tors but EXT-E. While low αs are no longer perceived as on many items, however, we suspect that some social de- indicators of low measurement quality [36–38], they do sirability or acquiescence bias might still be at play, sys- signal that our items cover different sub-aspects of the re- tematically inflating scores in relation to their “true spective dimensions rather than being extremely similar. values” for certain items. This warrants caution when This is no problem per se, but the relative conceptual interpreting absolute scores and calls into question the breadth of the motivation dimensions should be taken scale’s sensitivity “at the ceiling,” i.e., its ability to distin- into account when interpreting measurements. Should α guish respondents or measure change at high motivation be even lower in other settings, a re-evaluation of the scale levels. Generally, note that systematic biases are less of a items and the scale’s dimensionality might be necessary. concern when investigating relationships of motivation Second, factor correlations were relatively large in magni- with other variables or changes in motivation over time, tude compared to other SDT-based measures (e.g., [18]). assuming that biases stay constant. We believe there to be two main reasons: Respondents’ generally scored relatively high despite the various mea- sures in place, the common method and acquiescence bias Criterion validity likely inflating correlations [39]. Additionally, we found In addition to the convergent/discriminant validity ana- cross-loadings and residual correlations for many items, lyses in this study, it would be important to also examine Fig. 2 The modified SDT taxonomy of motivation as measured by the scale Lohmann et al. Human Resources for Health (2017) 15:33 Page 10 of 12 the scale against more tangible criteria such as work per- Use SEM for the actual analysis of interest formance in the future. Generally, substantive analyses on data collected with the scale can be done in one of the following two ways: One can either calculate composite scores or use them Recommendations for future use in any other type of analysis (e.g., predictor or outcome We welcome the use of the scale in future research and variables in regression models). Composite scores are are confident that the scale will prove a valid instrument usually calculated as the unweighted means of responses with health workers in other countries and settings as to all items pertaining to a factor/dimension. Alterna- well. The scale will be useful for researchers who want tively, one can continue in an SEM framework by adding to not only investigate overall levels of work motivation a structural part corresponding to a regression model to (“motivation intensity”) but also study how motivation the measurement model. The composite score calcula- of different origin and characteristics contribute to these tion is skipped and substantive relationships are directly overall levels (“motivation composition”) to understand estimated from the items via the latent factors, thus pre- how different “motivation profiles” relate to outcomes of serving full variance in the data. For this and other rea- interest [4]. sons, SEM is clearly preferred by psychometricians and Based on our experiences with the scale so far, we generally leads to better estimates [37, 42] but is statisti- would like to offer the following recommendations to re- cally complex and requires large samples [32]. searchers interested in using the tool: Beyond its general advantages, we also recommend SEM based on a number of specific results of our ana- lyses. Calculating composite scores bears a risk of impre- Use the full 26-item scale cision if systematic differences in item-factor loadings Use the full 26-item scale, if possible within the scope of (i.e., items have different indicative values for the motiv- your research. Although we are confident that the item ation factor) or intercepts (i.e., systematic differences in list covers the most important reasons for work motiv- item scores which are unrelated to the underlying motiv- ation even beyond Burkinabé nurses, our item selection ation level) are not accounted for. As with other biases, for the final 15-item scale was heavily empirically driven this is less of an issue if relationships between variables and thus reliant on the specific sample. We cannot ex- or change over time is the focus of interest, but of crit- clude that a different item selection would have resulted ical importance if interpretation of absolute motivation from a different sample. levels is planned. We found only slightly inhomogeneous factor loadings and intercepts in our sample which did Use a response scale with seven to nine options not seem to lead to substantial differences between com- Although our 11-point scale seemed to have had certain posite scores and latent factor scores. However, more advantages, we suspect that it might have overwhelmed substantial differences are possible in other settings. If some respondents, who might have had difficulty con- the use of SEM is not feasible, we strongly recommend ceptualizing the fine differences between scores on the developing a good understanding of all item properties 11-point scale. See Additional file 1 for a more extensive before embarking on substantive analyses with compos- discussion. ite scores. Should differences in factor loadings or inter- cepts across items be more substantial, one might consider weighing items when calculating composite Test for measurement invariance scores rather than giving equal weight to all items, or Test for measurement invariance to identify non- adding constants to balance differences in intercepts. invariant scale items before moving on to the actual ana- Note that such adjustments have implications for the lysis of interest. Our generalizability analyses suggest interpretation of the measurement (i.e., the “meaning” that it is possible to compare measurements for different and level of the respective motivation dimension), health worker subgroups on all statistical parameters depending on how each item effectively contributes to (e.g., means, variances) if analyses are performed in an the composite scores. They should thus be applied SEM framework. If factor means for different subgroups with caution. are to be compared using composite scores, however, systematic differences in scoring between groups are po- Conclusions tentially more problematic as they might artificially cre- This article presents evidence for the validity of a Self- ate non-real or mask real group differences [37, 41]. In Determination Theory-based scale to measure health our sample, respondents from different subgroups worker motivation composition. Our results show that showed somewhat different scoring behavior on items the scale measures a modified version of the SDT tax- im3, intro1, intro2, ext6, and ext7. onomy well and relatively consistently across health worker Lohmann et al. Human Resources for Health (2017) 15:33 Page 11 of 12 subgroups. Results of the convergent/discriminant valid- 2013-7-066). Written consent was obtained from all respondents prior to the ation indicate that the five dimensions of motivation relate survey. The database was anonymized to ensure respondents’ confidentiality. differently to important work outcomes, underlining the value of investigating motivation composition for the de- Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published velopment of a more profound understanding of health maps and institutional affiliations. worker motivation. We hope that our tool will contribute to meaningful research informing the design of effective Author details 1 Institute of Public Health, Faculty of Medicine, Heidelberg University, Im and side effect-free interventions to enhance motivation Neuenheimer Feld 324, Heidelberg, Germany. 2Centre de Recherche en and performance. Santé de Nouna, BP 02, Nouna, Burkina Faso. 3Département de Psychologie, Université du Québec à Montréal, C.P. 8888 succursale Centre-ville, Montréal, Québec H3C 3P8, Canada. 4World Bank, Health, Nutrition, Population Global Additional files Practice, 1818H Street, NW Washington, DC 20433, United States of America. 5 Département de Recherche Clinique, Centre MURAZ, 2054 Avenue Additional file 1: Development process of a new SDT-based motivation Mamadou Konaté, 01 BP 390, Bobo-Dioulasso, Burkina Faso. composition measure in Burkina Faso. (DOCX 287 kb) Received: 20 July 2016 Accepted: 11 May 2017 Additional file 2: Items eliminated in the analytical process, standardized parameter estimates for Model C, and suggested modification indices for Model C. (DOCX 107 kb) References 1. Campbell J, Dussault G, Buchan J, Pozo-Martin F, Guerra Arias M, Leone C, et al. A universal truth: no health without a workforce. Forum report, Third Abbreviations Global Forum on Human Resources for Health, Recife, Brazil. Geneva: Global AIC: Akaike information criterion; AUT: Autonomous motivation; CFI: Comparative Health Workforce Alliance and World Health Organization; 2013. fit index; CTRL: Controlled motivation; ESEM: Exploratory structural 2. Meessen B, Soucat A, Sekabaraga C. Performance-based financing. Just a equation modeling; EXT: External regulation; EXT-E: External regulation- donor fad or a catalyst towards comprehensive health-care reform? Bull economic subcomponent; EXT-S: External regulation-social subcomponent; World Health Organ. 2011;89:153–56. IDEN: Integrated/identified regulation; IM: Intrinsic motivation; INTRO: Introjected 3. Dieleman M, Gerretsen B, van der Wilt GJ. Human resource management regulation; LMIC: Low- and middle-income country; PBF: Performance-based interventions to improve health workers’ performance in low and middle financing; RMSEA: Root mean square error of approximation; RQ: Research income countries. A realist review. Health Res Policy Syst. 2009;7:7. question; SDT: Self-Determination Theory; SEM: Structural equation modeling; 4. Lohmann J, Houlfort N, De Allegri M. Crowding out or no crowding out? A SRMR: Standardized root mean square residual Self Determination Theory approach to health worker motivation in performance-based financing. Soc Sci Med. 2016;169:1. Acknowledgements 5. Renmans D, Holvoet N, Orach CG, Criel B. Opening the ‘black box’ of The authors would like to thank Andreas Neubauer for his support with the performance-based financing in low- and lower middle-income countries: a analysis and Danielle Wilhelm for her support in drafting this manuscript. review of the literature. Health Policy Plan. 2016. doi:10.1093/heapol/czw045. 6. Willis-Shattuck M, Bidwell P, Thomas S, Wyness L, Blaauw D, Ditlopo P. Funding Motivation and retention of health workers in developing countries. A This study was funded by the World Bank through the Health Results systematic review. BMC Health Serv Res. 2008;8:247. Innovation Trust Fund. JL received funding for her doctoral work from the 7. Bertone MP, Witter S. The complex remuneration of human resources for Konrad-Adenauer-Stiftung. We further acknowledge financial support by health in low-income settings: policy implications and a research agenda Deutsche Forschungsgemeinschaft and Ruprecht-Karls-Universität Heidelberg for designing effective financial incentives. Hum Resour Health. 2015;13:62. through the “Open Access Publishing” program. 8. Henderson LN, Tulloch J. Incentives for retaining and motivating health workers in Pacific and Asian countries. Hum Resour Health. 2008;6:18. Availability of data and materials 9. Pinder CC. Work motivation in organizational behavior. Hove: Psychology Data will not be made public at this point as the overall impact evaluation is Press; 2008. still ongoing. However, it is World Bank policy to make data generated in the 10. Franco LM, Bennett S, Kanfer R, Stubblebine P. Determinants and context of its projects public within 2 years of project completion. consequences of health worker motivation in hospitals in Jordan and Georgia. Soc Sci Med. 2004;58:343–55. Authors’ contributions 11. Bennett S, Franco LM, Kanfer R, Stubblebine P. The development of tools to The study was conceptualized by JL and MDA. JL, AS, JT, and MDA were measure the determinants and consequences of health worker motivation involved in the scale development process. JL, AS, PJR, and SS contributed to in developing countries. Bethesda: Abt. Associates; 2001. the main data collection. Analyses were performed by JL with support from 12. Prytherch H, Leshabari MT, Wiskow C, Aninanya GA, Kakoko DCV, Kagoné M, NH and MDA. JL drafted the manuscript with contribution from all authors. et al. The challenges of developing an instrument to assess health provider All authors read and approved the final manuscript. motivation at primary care level in rural Burkina Faso, Ghana and Tanzania. Glob Health Action. 2012;5:1–18. 13. Mbindyo PM, Blaauw D, Gilson L, English M. Developing a tool to measure Competing interests health worker motivation in district hospitals in Kenya. Hum Resour Health. The study was funded by the World Bank through the Health Results 2009;7:40. Innovation Trust Fund, of which PJR is a staff member. However, the World 14. Deci EL, Ryan RM. Intrinsic motivation and self-determination in human Bank did not interfere in the study design, analysis, or interpretation in any behavior. New York: Plenum; 1985. way. All other authors are not affiliated with the World Bank. 15. Gagné M, Deci EL. Self-determination theory and work motivation. J Organ Behav. 2005;26:331–62. Consent for publication 16. Deci EL, Ryan RM. The ‘what’ and ‘why’ of goal pursuits: human needs and Not applicable. the self-determination of behavior. Psychol Inq. 2000;11:227–68. 17. Miquelon P, Vallerand RJ. Goal motives, well-being, and physical health: an Ethics approval and consent to participate integrative model. Can Psychol. 2008;49:241–49. The study was approved by the Ethics Committee of the Faculty of Medicine 18. Gagné M, Forest J, Vansteenkiste M, Crevier-Braud L, van den Broeck A, Aspeli of Heidelberg University, Germany (protocol number S-272/2013), and by the AK, et al. The multidimensional work motivation scale: validation evidence in Ethics Committee of the Ministry of Health in Burkina Faso (protocol number seven languages and nine countries. Eur J Work Organ Psychol. 2015;24:178–96. Lohmann et al. Human Resources for Health (2017) 15:33 Page 12 of 12 19. Paul F. Health worker motivation and the role of performance based predicting employees’ satisfaction and turnover intentions. Eur J Work Organ finance systems in Africa. A qualitative study on health worker motivation Psychol. 2013;22:450–60. and the Rwandan performance based finance initiative in district hospitals. 44. van den Broeck A, Howard J, Leroy H, Gagné M. Motivated to work: a meta- Master’s thesis. London: London School of Economics; 2009. analysis of regulations. Conference presentation at the 6th International 20. Chandler CIR, Chonya S, Mtei F, Reyburn H, Whitty CJM. Motivation, money Conference on Self-Determination Theory, Victoria, Canada; 2016. and respect: a mixed-method study of Tanzanian non-physician clinicians. 45. Baruch-Feldman C, Brondolo E, Ben-Dayan D, Schwartz J. Sources of social Soc Sci Med. 2009;68:2078–88. support and burnout, job satisfaction and productivity. J Occup Health 21. Dieleman M, Cuong PV, Anh LV, Martineau T. Identifying factors for job Psychol. 2002;7:84–93. motivation of rural health workers in North Vietnam. Hum Resour Health. 46. Eisenberger R, Huntington R, Hutchison S, Sowa D. Perceived organizational 2003;1:10. support. J Appl Psychol. 1986;71:500–07. 22. Dieleman M, Toonen J, Touré H, Martineau T. The match between motivation 47. Boardman C, Sundquist E. Toward understanding work motivation. Worker and performance management of health sector workers in Mali. Hum Resour attitudes and the perception of effective public service. Am Rev Public Adm. Health. 2006;4:2. 2009;39:519–35. 23. Peters DH, Chakraborty S, Mahapatra P, Steinhardt L. Job satisfaction and motivation of health workers in public and private sectors: cross-sectional analysis from two Indian states. Hum Resour Health. 2010;8:27. 24. Goldberg AB, Ron LI. Understanding the complex drivers of intrinsic motivation for health workers in Malawi. Health Systems 20/20 project report. Abt Associates: Bethesda; 2012. 25. Morrison J, Batura N, Thapa R, Basnyat R, Skordis-Worrall J. Validating a tool to measure auxiliary nurse midwife and nurse motivation in rural Nepal. Hum Resour Health. 2015;13:30. 26. Ministère de la Santé du Burkina Faso. Annuaire Statistique 2014. Ouagadougou: Ministère de la Santé; 2015. 27. Codjia PL, Ouoba V. Motivation, leadership et performance des équipes dirigeantes du secteur public de la santé du Burkina Faso. Ouagadougou: Ministère de la Santé; 2003. 28. Scheewe S, Dieleman M, Millogo JJ, Traore A. Planification pour une couverture universelle: les ressources humaines en santé maternelle, néonatale et infantile au Burkina Faso en 2013-2025. Amsterdam and Ouagadougou: Royal Tropical Institute (KIT) and Société d’Études et de Recherche en Santé Publique (SERSAP); 2013. 29. Tremblay MA, Blanchard CM, Taylor S, Pelletier LG, Villeneuve M. Work Extrinsic and Intrinsic Motivation Scale. Its value for organizational psychology research. Can J Behav Sci. 2009;41:213–26. 30. Institute of Public Health Heidelberg University, The World Bank, Ministère de la Santé, Centre Muraz, Centre Hospitalier de l’Université de Montréal. Impact evaluation for health performance-based financing in Burkina Faso, Baseline Survey Report. Heidelberg: Institute of Public Health; 2015. 31. Vansteenkiste M, Lens W, De Witte S, De Witte H, Deci EL. The ‘why’ and ‘why not’ of job search behaviour: their relation to searching, unemployment experience, and well-being. Eur J Soc Psychol. 2004;34:345–63. 32. Kline RB. Principles and practice of structural equation modeling. 3rd revised edition. New York: Taylor & Francis; 2010. 33. Vandenberg RJ, Lance CE. A review and synthesis of the measurement invariance literature: suggestions, practices, and recommendations for organizational research. Organ Res Methods. 2000;3:4–70. 34. Bryant FB, Satorra A. Principles and practice of scaled difference chi-square testing. Struct Equ Model. 2012;19:372–98. 35. Fan X, Thompson B, Wang L. Effects of sample size, estimation method, and model specification on structural equation modeling fit indexes. Struct Equ Model. 1999;6:56–83. 36. Cortina JM. What is coefficient alpha? An examination of theory and applications. J Appl Psychol. 1993;78:98–104. 37. Borsboom D. The attack of the psychometricians. Psychometrika. 2006;71:425–40. 38. Sijtsma K. On the use, the misuse, and the very limited usefulness of Cronbach’s alpha. Psychometrika. 2009;74:107–20. 39. Podsakoff PM, MacKenzie SB, Lee JY, Podsakoff NP. Common method biases Submit your next manuscript to BioMed Central in behavioral research: a critical review of the literature and recommended and we will help you at every step: remedies. J Appl Psychol. 2003;88:879–903. 40. Marsh HW, Morin AJ, Parker PD, Kaur G. Exploratory structural equation • We accept pre-submission inquiries modeling: an integration of the best features of Exploratory and • Our selector tool helps you to find the most relevant journal Confirmatory Factor Analysis. Annu Rev Clin Psychol. 2014;10:85–110. • We provide round the clock customer support 41. Steinmetz H. Analyzing observed composite differences across groups: is partial measurement invariance enough? Meth Eur J Res Meth Behav Soc • Convenient online submission Sci. 2013;9:1–12. • Thorough peer review 42. Skrondal A, Laake P. Regression among factor scores. Psychometrika. 2001; • Inclusion in PubMed and all major indexing services 66:563–75. 43. Gillet N, Gagné M, Sauvagère S, Fouquereau E. The role of supervisor autonomy • Maximum visibility for your research support, organizational support, and autonomous and controlled motivation in Submit your manuscript at www.biomedcentral.com/submit