Motivation to test, treat, and report malaria cases: a quantitative assessment among private sector providers in the Greater Mekong Subregion

Background Accurately testing, treating, and tracking all malaria cases is critical to achieving elimination. Ensuring health providers are able and motivated to test, treat, and report cases is a necessary component of elimination programmes, and particularly challenging in low endemic settings where providers may not encounter a large volume of cases. This study aimed to understand provider motivations to test, treat, and report malaria cases to better optimize programme design, adjust incentive schemes, and ultimately improve reporting rates while growing the evidence base around private providers in the Greater Mekong Subregion (GMS). Methods With funding from the Bill & Melinda Gates Foundation, this study aimed to identify and validate distinctive subtypes of motivation among private sector providers enrolled in the Greater Mekong Subregion Elimination of Malaria through Surveillance (GEMS) programme, implemented by Population Services International. Quantitative questionnaires were administered electronically in person by trained enumerators to various provider groups in Myanmar, Lao PDR, and Vietnam. A three-stage confirmatory factor analysis was then conducted in STATA. Results Following this analysis, a two-factor solution that describes motivation in this population of providers was identified, and providers were scored on the two dimensions of motivation. The correlation between the two rotated factors was 0.3889, and the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was 0.93, indicating an excellent level of suitability. These providers, who are often assumed to only be financially motivated, engaged in malaria elimination activities because of both internal and external motivational factors that are independent of remuneration or financial gain. For all three countries’ data, significant covariances between the two latent variables for internal and external motivation were found. The models were found to be of adequate to good fit for the data across all three countries. It was determined that private sector providers, who were previously believed to be primarily financially motivated, were also motivated by personal factors. Motivation was also associated with key outcomes of importance to malaria elimination, such as reporting and stocking of tests and treatments. Conclusion Maintaining or increasing provider motivation to test and treat is essential in the fight to eliminate malaria from the GMS, as it helps to ensure that providers continue to pursue this goal, even in a low incidence environment where cases may be rare and in which providers face financial pressure to focus on areas of health service provision. Establishing mechanisms to better motivate providers through intrinsic factors is likely to have a substantive impact on the sustainability of malaria case management activities.


Background
Important gains towards malaria elimination have been achieved in recent years in the Greater Mekong Subregion (GMS). Although malaria cases and deaths have declined dramatically, the expansion of artemisinin resistance in the region is a growing threat to malaria control efforts. Thorough surveillance is critical to curbing the epidemic, and private sector contributions, while important, often go unreported [1,2]. Established in 2016 and funded by the Bill and Melinda Gates Foundation, the Population Services International (PSI) Greater Mekong Subregion Elimination of Malaria through Surveillance (GEMS) programme aims to increase private sector engagement to accelerate progress towards elimination.
Motivation has been defined as, "the level of effort and desire to perform well" and is an important determinant of quality of care [3]. Motivation in an employment setting is defined as "… a set of energetic forces that originate both within as well as beyond an individual's being, to initiate work related behavior, and to determine its form, direction, intensity, and duration" [4]. Motivation has been associated with lower levels of staff turnover [5], higher retention, less job burnout, increased performance [6], and higher quality of care [7]. Further, motivated employees come to work more regularly, work more diligently, and are more flexible [8].
The evidence base on provider motivation in a malaria context is limited, particularly in terms of a robust, multicountry quantitative analysis. Existing studies tend to be qualitative in nature, conducted in the public, not private, sector, and largely focused on community health workers (CHWs) [9][10][11]. Within the malaria context, existing literature explores provider motivations for testing and treatment [12][13][14], but little is known regarding provider motivations for case reporting. There is also a dearth of literature for Plasmodium falciparum elimination contexts [15,16].
Quantitative measurement of provider motivation involves defining motivation, a multidimensional construct. It must also consider the multiple components of motivation that influence behavior, and the contextspecific language used to discuss motivation in different cultural settings. Furthermore, employment motivations can differ in both conceptualization and measurement between different provider subgroups. Direct measures are typically derived through measurement scales within a survey or through qualitative methods. Indirect measures of motivation can be derived through surveys or qualitative methods via experimental games or observations of behaviour [17]. Further, where measuring motivation has been attempted, it has been focused on the public and community sectors.
In 2000, Bennett, Franco, Kanfer and Stubblebine developed a specific tool to measure the determinants and consequences of public sector health worker motivation in developing countries, which was then used in a three-part study of health worker motivation in hospitals in Jordan and Georgia [18]. The tool encompasses a number of motivational and performance categories, theoretical constructs, and scales.
In 2017, Lohmann et al. [19] developed a psychometric scale to measure motivation composition. The scale was grounded in the self-determination theory (SDT), a theory introduced in the 1980's as a general framework of human motivation. The SDT captures a generalized measure of motivation toward work and identifies five dimensions of motivation that can be placed along a continuum from extrinsic (motivation to attain or avoid a consequence that is maintained by rewards/punishment) to intrinsic (motivation stemming from the enjoyment of a task).
Much progress has been made in the GMS, particularly during the last five years, to reduce the malaria burden. In Lao PDR, the number of confirmed malaria cases decreased between 2012 and 2019 by 80% [20]. Vietnam has fewer than 5000 confirmed cases per year, mostly concentrated in three provinces: Binh Phuoc, Dak Lak, and Gia Lai [21]. In the past decade, the number of reported malaria deaths in Myanmar has dropped steadily year by year from 1707 in 2005 to just 19 in 2018 (a 99% reduction over 10 years). The incidence of reported malaria has fallen by 85% since 2012 (from 9.94 per 1000 population in 2012 to 1.46 per 1000 population in 2018) [20].
Engaging the private sector is necessary to achieve malaria elimination in the GMS, as a significant proportion of the population first seeks health care within that sector. The preference for private sector providers is likely related to accessibility and perceptions of quality and flexibility in prescribing medicines when compared to the public sector [22]. However, none of the six countries environment where cases may be rare and in which providers face financial pressure to focus on areas of health service provision. Establishing mechanisms to better motivate providers through intrinsic factors is likely to have a substantive impact on the sustainability of malaria case management activities. Keywords: Malaria, Malaria elimination, Private sector, Provider motivation, Confirmatory factor analysis in the GMS collect complete case data from private sector points of care such as pharmacies, clinics, shops and private hospitals [23]. As a result, national policy makers lack access to a complete malaria case data set to inform programme strategies and interventions.
In 2015 and 2016, cross sectional outlet surveys identified low availability of malaria diagnostic testing (Cambodia, 75%; Lao PDR, 94%; Myanmar, 75%) and poor access to first-line treatment in the private sector (Cambodia, 70.9%; Lao PDR, 40.8%; Myanmar P. falciparum = 42.7%, P. vivax = 19.6%) across the GMS [24,25]. Between 2015 and 2019, the GEMS programme received funding from the Bill & Melinda Gates Foundation to support national malaria control programmes (NMCPs) in Cambodia, Lao PDR, Myanmar and Vietnam to capture private sector data by engaging private providers in malaria case management, generating private sector malaria case data, and integrating these data into national surveillance systems. The GEMS established network has increased access to quality case management in Cambodia, Lao PDR, Vietnam and Myanmar, where use of the private sector remains high [22]. PSI-supported providers, funded from the GEMS project, detected between 1.8 and 18% of the national caseload in each of the four countries in 2019 [25].
Achieving malaria elimination will require a concerted effort from actors across the public and private sectors. Private sector involvement in malaria surveillance is a relatively recent development. This study sought to understand the motivations of private sector network providers to test, treat, and report malaria cases and determine barriers to quality performance. Quality performance within the GEMS programme defined as achieving 80% or more during quality assessment visits. The quality of care is monitored routinely using standardized checklists developed in collaboration with the national programme and according to international benchmarks for quality malaria service provision. The present study aims to identify and validate different underlying subtypes of motivation among private sector providers enrolled in the GEMS programme by using a three-stage confirmatory factor analysis followed by a two-factor solution. Similar modelling approaches have previously been applied to understand determinants and barriers to vaccine coverage [26] and malaria control measures [27]. How these subtypes of motivation differ by provider characteristics was then examined, in addition to whether they are associated with intentions and outcomes related to malaria service provision. These analyses will allow us to develop recommendations that influence policies on the role of the private sector in national malaria elimination strategies.

Study population
Formal healthcare providers (for the purposes of the study are providers supported by the GEMS project and permitted to operate in the national context) served as the sampling frame for this study and were eligible for inclusion if they were actively enrolled in the GEMS malaria programmes and consented to participate. GEMS works with different provider types in each country, thus specific provider types varied accordingly.
In Myanmar, three types of providers were sampled from the PSI supported networks: POs, ICMVs, and SUN Network providers. In Vietnam, three provider cadres were targeted for this study: clinics, pharmacies, and FMCGs (CMCs were excluded from this study). In Lao PDR, clinics and pharmacies were included in this study. In Lao PDR, these providers are largely identical and have similar educational and training requirements and responsibilities.
A random sample of providers, stratified by type of provider, was drawn from each country's list of providers who met the study inclusion criteria (Table 1). A sample size calculation for the number of providers needed to estimate key measures was used with a precision of ± 7.5%. Providers within each stratified group were selected using a simple random sampling. In Myanmar, providers were selected using systematic sampling.

Survey instruments
For this study, a quantitative survey instrument was developed, borrowing from previously tested measures of motivation [18,19] and incorporating additional questions with input from in-country expert teams with the dual aim of ensuring programmatic relevance while generating robust measures. Respondents were asked a range of questions about their motivations for participating in PSI's malaria programme, measured using five-point Likert-type scales. All motivation questions were identically written and administered across the three countries, while some programmatic questions varied.
Questionnaires were administered in person by trained enumerators and responses were captured electronically. This study was approved by the PSI Research and Ethics Board and local review boards in each of the three countries (Myanmar MM-PSI REB #26.2018; Lao PDR LA-Local IRB #2018.69.MP and Vietnam VN-Local IRB #462/2018/YTCC-HD3.) All participants provided informed consent and data were deidentified prior to analysis. The provider motivation module is shown in Annex 1.

Analysis
Provider motivation was conceptualized to be a multidimensional construct and, given the dearth of previous research into provider motivation in the GMS, a large number of survey items hypothesized to relate to provider motivation were included in the data collection tool. To first establish that the items used in the questionnaire did in fact pertain to the different elements of motivation suggested in the literature, [17] an exploratory factor analysis (EFA) was conducted followed by a review of the substantive content of the questions to validate that the questionnaire items were measuring two constructs, intrinsic and extrinsic motivation, in the Myanmar data. Next, confirmatory factor analysis (CFA) was used to test construct validity in the Lao PDR and Vietnam data. Factor loadings were then used to calculate weighted scores, which were used in subsequent analyses. The analysis proceeded in the following stages: Stage 1: The analysis was conducted initially on the Myanmar dataset. EFA was used to explore the underlying structure of the correlations between the survey items and to develop a parsimonious set of provider motivation questions relevant to this context. A two-factor model was fitted based on an examination of scree plots and factor eigenvalues. Of the original 32 items pertaining to motivation in the dataset, 16 items with communalities greater than 0.5 were retained. Using maximum likelihood estimation, and promax (oblique) rotation of the factors, a simple solution was achieved (i.e. each item loading onto only one factor). The use of oblique rotation reflects our expectation that the two motivation subtypes are correlated.
Stage 2: A CFA model was developed, using the structural equation modelling (SEM) command in Stata, with two latent variables and item loadings reflecting the simple solution found in the EFA model. This model was first fitted for the Myanmar dataset. The same model specification was then applied separately to the Lao PDR and Vietnam datasets to validate the proposed structure. The results and Satorra-Bentler (adjusted for non-normality) goodness of fit statistics are detailed in the results section below.
Stage 3: For each of the three CFA models (Myanmar, Lao PDR and Vietnam), predicted Bayes scores were calculated (with zero mean, and unit variance) for each provider on both latent factors using the SEM postestimation predict, latent command in STATA. Further analyses were conducted on these scores to examine how these two dimensions of provider motivation were associated with a series of background demographics and outcome variables of interest. Significance testing was performed on these analyses using oneway ANOVA, simple linear regression, or t-tests according to variable type.

Results
Respondents were majority female (53% Myanmar, 77% Lao PDR, and 54% Vietnam), averaging 44, 48, and 43 years of age in Myanmar, Lao PDR, and Vietnam, respectively ( Table 2). In Myanmar respondents had an average of 5 years of experience working as a provider, whereas in Vietnam providers had an average of 15 years of experience. In Lao PDR, this question was not asked. Instead, providers were asked how long they had worked in the PSI programme, with an average of 6 months reported. Levels of education varied widely across countries.

Factor analysis
Following the three-stage process described above using data from Myanmar, EFA identified a two-factor solution that describes motivation in this population of providers. Annex 3 shows the unrotated and rotated factor loadings for a bidimensional model for provider motivation with loadings greater than 0.4 indicated in bold. The correlation between the two rotated factors was 0.3889, and the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy was 0.93, indicating an excellent level of suitability.
The face validity of the two factor solution rests on the theoretical model for motivation that identifies internal and external motivation as key constructs. Examining the questionnaire items that are associated with each of the two factors in the EFA confirms that internal and external motivation are the two key dimensions of provider motivation in Myanmar. The positive correlation between internal and external motivation dimensions confirms our assumption that both represent different dimensions a larger concept of "provider motivation" in this context.
As the second stage in the analysis, a confirmatory model was estimated for the Myanmar data (Model 1, Table 3). Figure 1 below shows this model. The same bidimensional structure was then applied to the Lao PDR (model 2) and Vietnam (model 3) datasets. The model coefficients, P-values, covariance between latent variables, and goodness of fit statistics are shown in Table 3. Recognizing that financial motivations were predicted to be important in the literature, a further confirmatory factor analysis was conducted, adding a third latent construct for financial motivation, measured by three financial motivation-related items in the questionnaire. These items were only asked to two of the three provider types in Myanmar who receive financial compensation (model 4, Table 3).

Loadings and country comparison
Four variables constitute the construct of external motivation in the original Myanmar analysis (model 1), and this was replicable for the Vietnam dataset (model 2), with all coefficients of a similar magnitude and statistical significance in both countries. In applying this model structure to the Lao PDR provider data (model 3), however, it was determined that one coefficient for the statement, "Because my reputation depends on it" was not significantly associated with the latent construct. Furthermore, a second coefficient for "It brings pride to my family to know that I'm contributing to malaria elimination" had a small loading. This suggests that the external motivation construct may look different or was not fully captured for providers in Lao PDR. The latent construct of internal motivation, modelled originally in the Myanmar dataset, was stable across both additional countries, with statistically significant coefficients of a comparable magnitude for Lao PDR and Vietnam. For all three countries' data, positive, significant covariances between the two latent variables for internal and external motivation were found. The covariance was higher in Lao PDR and Vietnam than in Myanmar.

Goodness of fit
Various measures of goodness of fit for structural equation and CFA models are proposed in the literature. Several are reported here in line with recommended best practices in SEM modelling. The models were found to be of adequate to good fit for the data across all three countries. The chi-squared test statistic, while indicative of good fit, is also sensitive to violations of its assumptions and may not be a very good indicator of model adequacy, particularly with smaller sample sizes [28]. To correct for sample size sensitivity, the chi-squared to degrees of freedom ration may additionally be considered. Values below 3 indicate model adequacy, with lower values indicating better fit. All three models presented here have adequate fits according to this measure. RMSEA is less sensitive to  sample size than chi-squared. Values below 0.05 indicate close fit, between 0.05 and 0.08 fair fit, and between 0.08 and 0.1 mediocre fit [29]. Comparative fit index (CFI) values greater than 0.9 are considered to indicate good fit.
In this analysis, both Lao PDR and Myanmar attain this criterion. Finally, the standardized root mean squared residual (SRMR) is a standardized measure of the difference between observed and predicted correlations, with a value of less than 0.08 indicating good fit. Models for both Lao PDR and Myanmar achieved this level, with Vietnam exceeding the cutoff only slightly.
To understand whether provider characteristics are associated with different levels of internal and external motivation, scores were predicted on the two dimensions of motivation for all providers in the three country's datasets (Tables 4, 5, 6). These predicted scores were then used in subsequent analyses to understand how the two motivation subtypes vary according to characteristics of providers.
Predicted scores for both latent constructs of motivation were standardized with zero mean and unit variance. Female providers in Myanmar had significantly higher average scores on the internal motivation scale than their male peers, while the opposite was found in Vietnam. No difference was found between female and male providers in Lao PDR.
Some geographic variation in motivation across all three countries was seen. External motivation in Myanmar was highest among providers in Tanintharyi Region and lowest in Sagaing Region, while internal motivation was highest in Shan State and lowest in Mandalay Region. In Lao PDR, providers in Saravan and Savannakhet scored highest on both internal and external motivation scales, while those in Attapeu and Sekong scored lowest. In Vietnam, significant differences by region were only seen for the internal motivation scale, where providers in Binh Phuoc scored highest, and those in Gia Lai lowest.
The education level of providers was significantly related to both dimensions of motivation in Myanmar and Vietnam, but the direction of effect differed between the two countries. In Myanmar, higher levels of education were associated with lower levels of both subscales of motivation. In Vietnam, providers reporting more advanced levels of education scored higher on both subscales of motivation.
The age of providers was significantly related to external and internal motivation scores in both Myanmar and Vietnam, with increased provider age associated with a small but significant increase in the standardized external motivation score in both countries, while internal motivation decreased with age in Myanmar and increased with age in Vietnam.
The measure of time spent in the PSI malaria programme was significantly related to providers' internal motivation scores in Myanmar and Vietnam. In Myanmar, each additional year in the programme was associated with a small but significant decrease in internal motivation, while in Vietnam the relationship was positive.
The study also sought to understand what role external and internal motivation might play in predicting some key outcomes. In Myanmar and Vietnam, higher internal and external motivation scores were significantly associated with greater willingness to continue to maintain and share records with the government after the end of programme implementation. This implies that providers with higher motivation would continue to test, treat, and report malaria cases once financial incentives provided by the programme are withdrawn. This question was phrased differently in Lao PDR, where providers were asked if they would be willing to continue to share malaria case records by phone or SMS after the end of the programme. Here also for both internal and external motivation, significantly higher scores were associated with affirmative responses.  Finally, providers were asked what they would do to maintain malaria test and treatment stocks if the programme ended. In Myanmar those reporting that they would not restock had significantly lower levels of internal motivation than those with any other response. In Vietnam those reporting that they would not restock were found to have significantly lower scores on both internal and external motivation.
The inclusion of a third type of motivation related to financial benefits of being involved in the malaria programme resulted in a CFA model with adequate fit to the data (model 4 in Table 3, and Fig. 2). The coefficients and P-values for items loading on to internal and external motivation remained similar to the two-factor model (Models 4 and 1 in Table 3, respectively). The three finance-related items loaded positively on to the latent variable for financial motivation and all loadings were statistically significant at the 1% level. The structural part of the model suggests that when financial motivation is included, the positive and significant relationship between internal and external motivation persists. A non-negligible positive, statistically significant covariance Table 5 Myanmar predicted external and internal motivation scores Bold scores indicate significant difference within group at the 5% level (ANOVA); significance of regression coefficients = * = 5%; ** = 1%; *** < 0.1% between external and financial motivation types was also found. This means that providers who reported being more strongly motivated by external factors were also likely to be more strongly motivated by financial factors.

Myanmar
The relationship between internal motivation and financial motivation was not statistically significant. The predicted factor scores for the three motivation constructs (including financial motivation) are shown Table 6 Lao PDR predicted external and internal motivation scores Bold scores indicate significant difference within group at the 5% level (ANOVA); significance of regression coefficients = * = 5%; ** = 1%; *** < 0.1%   Table 7. There was little statistically significant variation in financial motivation by background demographics. This suggests that the degree to which providers are motivated by money in their malaria testing and treatment activities is less related to age, education or gender than is the case for internal and external motivation. Likewise, financial motivation scores were not significantly related to the outcome of intention to continue to maintain records after the end of the PSI programmeone measure of programme sustainability. However, for the other outcome of source of supplies after the end of the programme, there was a significant relationship with financial motivation: providers who reported that they would not restock testing and treatments scored significantly higher on the financial motivation scale than their peers.

Discussion
Ensuring that every suspected malaria case presenting to the public or private sector is tested and treated is critical to malaria elimination. To achieve this, motivating private providers to engage in national malaria response is key, particularly in areas where a significant proportion of the population seeks care in the private sector. The measurement of healthcare provider motivation is difficult because it is a transitory construct that can be unidimensional or multidimensional and can be measured directly or indirectly [17]. Existing scales that Table 7 Vietnam predicted external and internal motivation scores Bold scores indicate significant difference within group at the 5% level (ANOVA); significance of regression coefficients = * = 5%; ** = 1%; *** < 0.1%

Vietnam
Mean scores

External motivation Internal motivation
Provider characteristics  [30]. These measures have mostly been developed for use in highincome countries and may not be appropriate for use context lower-and-middle income countries. This study applied previous measures of motivation to a diverse group of private providers in three countries in a malaria elimination context. The measures were developed from existing, validated tools developed to measure motivation of health care providers in low-and middle-income countries (LMICs). The scale developed by Lohman and colleagues in 2017 [19] is based on SDT and measures motivation composition (the relative contribution of different kinds of motivation to overall work motivation). SDT is recommended for supporting programmes to determine how motivation of different origins and characteristics contributes to overall motivation and to understand how differences in the dimensions of motivation are associated with outcomes of interest. Also included were measures from Bennet and colleagues [18], who developed scales drawing from published literature for constructs of motivational determinants, such as worker expectations, values/work ethic, work-related personality, and emotional personality in Jordan and Georgia. These measures were adapted for use within the GEMS malaria elimination programme in Vietnam, Myanmar, and Lao PDR. These adaptations included slight wording changes and removal of redundant items, which were later pretested prior to study implementation. As found by other research in different contexts [31], the GEMS private sector providers, who were believed to be financially motivated, were also motivated by other personal factors. Motivations varied by key characteristics of providers and were predictive of outcomes of importance. Maintaining or increasing provider motivation to test and treat in malaria case management is essential in the fight to eliminate malaria from the GMS, as it helps to ensure that providers continue to pursue this goal, even in a low incidence environment where cases may be rare and in which providers face financial pressure to focus on areas of health service provision other than malaria case management.
These results demonstrate that in Myanmar and Vietnam (and to some extent in Lao PDR), provider motivation has two similar dimensions across countries. The relationship between internal and external motivation was also similar in all 3 countries, meaning that across these contexts, providers who have a higher level of external motivation are also likely to have a higher level of internal motivation. Examining how motivation varies by provider characteristics, results were highly countryspecific. This is unsurprising-Vietnam, Lao PDR and Myanmar are very different contexts, with differently structured health systems and incentive systems for providers. While recent decades of underinvestment have weakened Myanmar's public health system, significant improvements have been made during the last 5 years. Myanmar remains the most permissive environment for the private sector, and all PSI providers-including informal private outlets and volunteers-are able to test and treat Plasmodium falciparum and Plasmodium vivax malaria, including prescribing primaquine for radical cure of P. vivax.
Conversely, only private clinics are allowed to test and treat for malaria in Vietnam. In agreement with some provincial authorities, however, PSI has trained pharmacies, CMCs, and FMCG shops to test and refer malaria cases. These country-level differences may be reflective of the interplay between individual provider-level motivations and a unique relationship between organizational structure, culture, and societal culture [18].
When several important outcome variables in the analysis (willingness to continue to report cases after the end of the PSI malaria programme, and intention to acquire testing and treatment stocks post-PSI programme), were examined, it was found that in both Myanmar and Vietnam, higher provider internal and external motivation scores were associated with greater intent to continue these key aspects of malaria case management. This finding is timely as the programme phases out and the private sector provider networks transition to public sector oversight in each country. Finding ways to better motivate providers through intrinsic factors is, therefore likely to have a substantive impact on the sustainability of these activities during and after the implementation of this programme. It is thus possible to speculate that a fruitful approach to driving greater project sustainability may therefore lie in targeted recruitment or stratifying providers within each project country and developing interventions that will appeal to their core motivations. This challenges the conventional wisdom that providers only care about money and financial incentives are their primary source of motivation [32]. Indeed the analysis exploring financial motivation suggests it may only be quite weakly associated with external motivation in the model. Further research is needed to assess the degree to which these different dimensions of provider motivation may be change through time and, therefore, would have the potential to be increased through specific interventions. Similarly, it would be beneficial to know the degree to which this bidimensional structure of provider motivations might apply to other malaria service providers outside of the GMS. Further research should aim to identify interventions that increase internal and external motivations and to better understand the interplay between financial, internal, and external motivations and the value for money associated with interventions.

Limitations
The findings may not be generalizable to the private sector as a whole, as the study sample consisted only of private providers that engaged in the PSI GEMS programme. Further, these results would need to be validated for use in other contexts and provider types. Participation in the PSI programme could have led to response bias among study participants. Self-report bias could also have influenced results. The analysis was based on Likert-scale responses, which are imperfect and treated as continuous variables, an approach commonly used in the literature [17], despite being discrete. Likewise, the use of Likert scale-type questions may violate the assumption of normality under a maximum likelihood estimation method for the CFA models. For the financial analysis, the model was only an adequate fit, perhaps because the sample size was smaller and the model was more complex. The findings are indicative but should be interpreted with caution. Finally, the providers' reported outcomes were based on intentions, not observed actions, and further research might consider a longitudinal approach for examining associations between provider motivation and outcomes.

Conclusions
Conventional wisdom has long held that private sector providers are primarily motivated by financial incentives. These results, however, show motivation to be multifaceted for this group. This study consistently identified two dimensions: internal and external motivation, across three contexts and different types of providers (ranging from informal to medical professionals). Providers chose to join PSI's malaria programme for a variety of reasons, including commitment to serving their community, boosting their reputation, having access to professional development opportunities, and receiving commodities.
By understanding how motivation varies according to provider characteristics, malaria elimination programmes can better target continuing professional development, adapt incentive structures, and update training and routine communication with providers to build on factors that may improve internal and external motivation. These findings provide national programmes with the opportunity to better understand providers within their contexts which in turn can lead to better programme design leveraging appropriate incentive schemes and motivations in order to enhance provider performance and programme results in the local context.
Organizations need to look at providers not just as business owners and public servants, but also as nuanced actors with multiple sources of motivation. Individual characteristics are important to how private providers should be trained, recruited, and engaged to ensure long term success and sustainability. Future research should aim to better understand how motivation varies in different contexts and its effects on outcomes in the health system.  [25]. The SUN providers detected 13.4% of all PSI networks' positive cases, the ICMV channel detected 57.9%, and the PO channel detected 28.6%. Despite receiving fewer performance-based incentives (see Table  8), the PO network tends to have the highest test positivity rate. SUN doctors have the second highest (despite lower testing rates than ICMVs and POs), likely due to their urban and peri-urban location and qualifications, which makes them more likely to test for confirmation of clinical diagnosis. In Lao PDR, the PSI program operates in the five southern provinces and the low-burden, eliminationtargeted north. In 2019 PSI supported 474 public-private mix (PPM) providers, consisting primarily of doctors and pharmacists, to test for, treat and report uncomplicated Pf and Pv cases (with ACT only). For their participation in the GEMS program, the PPM providers in Lao PDR receive a maximum incentive of $40 per month intended to cover things such as internet costs. In 2019, GEMS supported providers tested 73,754 fevers (13% of total fevers tested in the country) and detected 612 cases. A total of 9.2% of the national reported caseload in Lao PDR was detected through the PSI PPM network.

Annex 1: Provider Motivation Questions
In Vietnam, the GEMS program operates in 4 provinces, primarily in the Central Highlands region. The GEMS network in Vietnam consists of private clinics (staffed by medical doctors), private pharmacies, community-based volunteers known as community malaria champions (CMCs), and fast-moving consumer goods shops (FMCGs). In agreement with some provincial authorities, PSI has trained pharmacies, CMCs, and consumer good shops to test and refer malaria cases. Providers in Vietnam do not receive financial compensation-only monthly/quarterly promotional material incentives valued at less than $43. In 2019, PSI's 828 GEMS supported providers tested 28,421 fevers and detected 877 cases, accounting for 1.4% of the total fevers tested in country and 18.7% of the national reported caseload. The majority of cases in the PSI network Table 8 Myanmar predicted external, internal, and financial motivation scores Bold scores indicate significant difference within group at the 5% level (ANOVA); significance of regression coefficients = * = 5%; ** = 1%; *** < 0.1%