Developing Sustainable Healthcare Systems in Developing Countries: Examining the Role of Barriers, Enablers and Drivers on Knowledge Management Adoption

: Knowledge management (KM) adoption is crucial to integrating sustainable development within the healthcare sector. Different barriers, enablers, and drivers affect KM adoption. Identifying these barriers, enablers, and drivers and their role in KM adoption is the core of successful KM adoption. However, there is scarcity of studies applying quantitative models and combing barriers, enablers and drivers to check their effect on KM adoption, especially form a developing country’s perspective such as Pakistan. Therefore, this study explores the role of barriers, enablers and drivers on KM adoption in Pakistan. Healthcare professionals participated in the data collection process, and results were analyzed using structural equation modeling. The ﬁndings described that: (1) organizational and strategic barriers have signiﬁcant negative association with KM adoption; (2) government related enablers have signiﬁcant positive association with KM adoption; (3) healthcare related drivers, and performance-based drivers have signiﬁcant positive association with KM adoption. This study concludes that government intervention to promote KM adoption is necessary especially in developing countries. These ﬁndings will be helpful for the healthcare professionals and policy makers to promote KM adoption in healthcare sector. Current study contributes to the healthcare literature and body of knowledge by providing the empirical evidence of checking the quantitative effect of barriers, enablers and drivers on KM adoption.


Introduction
Sustainability as an emerging issue and has been widely discussed in the healthcare sector. Considerable literature focused on the need of the sustainable, efficient, and effective healthcare [1]. However, currently, the healthcare systems are facing multiple challenges to cope with the current healthcare needs, and sustainability is considered to be a significant requirement to obtain strategic a fit for the future [2,3]. The world business council for sustainable development (WBCSD) defined

The Need for KM in the Healthcare of Pakistan
Pakistan is situated in South Asia; it has a population of over 212 million and a growth rate of 2% [35], making it the 6th most populous country in the world. It has an annual budget of PKRS 5,246.2 billion and is growing by 5% annually. GoP spent PKRS 13,897 million on its healthcare in 2017-2018 fiscal year, which is about 0.75% of the budget, and the percentage amount has steadily increased over the years [36]. The GoP is under the impression that by increasing the budget spending on healthcare it can improve the service of healthcare, but despite this the healthcare service is not improving.
The public sector hospitals in the healthcare sector take up most of the budget. The public sector hospitals of developing countries use up a lot of resources, like money and trained personnel. It is estimated to be 50-80% varying on the country [37]. The healthcare organizations have also become rigid and are having trouble adjusting to the rapid changes in the healthcare sector globally, and they are unable to provide quality healthcare to their customers as per their desires [38]. The healthcare of Pakistan currently ranks 149 out of 188 among the United Nations (UN) member countries in terms of healthcare goals [39]. The healthcare of Pakistan, despite growing, has always been under pressure due to disease outbreaks, natural disasters, large amount of information available on internet, and alternate healthcare delivery systems [40]. Pakistan is suffering from many diseases, the major diseases being neonatal disorders 20.4%, cancer 7%, ischemic heart disease 6.4%, lower respiratory infections 4.94%, stroke 3.42%, chronic kidney disease 1.45%, malaria 0.43%, etc. (figures according to [41]). Other than diseases, there are people that suffer due to disasters (earthquake and floods) and terrorism. During the earthquake of 8th October 2005 there was chaos; there were many patients pouring in and most of them in critical condition [42]. All the patients had to go through many tests to get their details, life prevalent conditions, blood type etc. This resulted in loss of time, increase in cost, and loss of life. If there was an effective knowledge management system these problems could have been overcome.
The GoP has now realized that increasing the budget is not a solution to the problem, they need to look for new methods. They are now considering the adoption of KM in their healthcare. It helps in the effective utilization of resources, adoption of best practices, rapid response to change and creating a competitive advantage [3]. KM helps with the storing and sharing of knowledge. If a new patient comes to a doctor with an improper record of health, then there is a chance of improper treatment and wastage of doctor's time increasing the cost, as research by Hersch W. R. [43] shows that improper documentation takes up 1/3 of the doctor's time. KM also helps the healthcare professional keep updated with the latest knowledge. Generally it is not possible for a doctor to keep up with new knowledge because there are over 200,000 medical journals, with over 7000 types of prescriptions, 800 tests, 1000 image tests, 1500 surgical procedures [8].

Knowledge Management Barriers to Healthcare
There have been several studies regarding barriers to KM. Singh, et al. [11] addressed unsupportive organizational culture, lack of leadership, improper strategic planning, lack of knowledge resources, lack of financial resources, improper technological infrastructure, lack of innovation and knowledge creation, integration of system, and inability to capture information as the main barriers to KM. Karamitri, et al. [44] identified several barriers to KM, such as slow transfer of information, sharing of useless information, inaccurate information, information overflow, and lack of time with a physician to keep updated. Hojabri, et al. [18] mentioned eight barriers; improper process activities, no training and education, absence of performance checks, ineffective knowledge management strategies, insufficient technology, a non-supporting organizational culture, and lack of management and leadership support. Kothari, et al. [45] stated the barriers under two categories; organizational level and individual level. The organizational barriers are the goals of KM conflicting with those of employees, the frequent turnover of employees, improper monetary reward system, no standard definition of KM, the tasks are not clear, lack of balance between information technology (IT) and people, the cost of implementing KM, the uncertainty of successful implementation, and unsupportive structure and culture. The individual barriers are lack of motivation, resistance to change, insufficient technological knowledge, unqualified authority, unwilling to share information, and employee turnover. The barriers identified by Sharma and Singh [9] are lack of top management commitment, unsupportive structure, high turnover, lack of knowledge about technology, no learning from past mistakes, reluctance to use technology, difference between company and individual goals, lack of trust among the employees, improper training, unavailability of time, restriction on the flow of knowledge, improper reward system, unsupportive organizational culture, lack of integration of KM, and lack of financial resources.
The knowledge management barriers to healthcare are given in Table 1. Fear of sharing incorrect information [51,77,78] The literature shows that barriers make it difficult for the stakeholders to adopt KM in their healthcare; the barriers have a negative influence. Due to this the current study considers the following hypotheses regarding the barriers: Hypothesis 1a (H1a). Organizational barriers have negative influence on the adoption of KM in healthcare.

Hypothesis 1b (H1b). Strategic barriers have negative influence on the adoption of KM in healthcare.
Hypothesis 1c (H1c). Technological barriers have negative influence on the adoption of KM in healthcare.

Hypothesis 1d (H1d). Resource barriers have negative influence on the adoption of KM in healthcare.
Hypothesis 1e (H1e). Individual barriers have negative influence on the adoption of KM in healthcare.

Knowledge Management Enablers of Healthcare
After a detailed literature review, several researches regarding the enablers of KM were studied. Pee and Kankanhalli [79] identified KM technology, supportive organizational structure, senior management championship, good social capital, high KM capability, and good organizational effectiveness to be the most important enablers for KM implementation in healthcare. Yeh, et al. [13] stated four main enablers to KM; effective information technology infrastructure, motivated employees, good corporate culture and strategy, and leadership. Lee and Choi [14] mentioned supportive organizational culture and structure, people, and technology. Karamitri, et al. [44] revealed IT, leadership, quick knowledge sharing, proper workflow assignment, elimination of distrust, open communication channels, motivation of employees, knowledge brokers and willingness of employees to share information as critical enablers of knowledge management. Kothari, et al. [45] identified training and education, a proper framework for knowledge management implementation, Sustainability 2019, 11, 954 6 of 31 ability to identify a knowledge broker, support from the management, organizational structure and culture as the main enablers of KM.
The knowledge management enablers of healthcare are given in Table 2.

ERE4
Training and education [18,19,48,99] The literature shows that enablers make it easy for the stakeholders to adopt KM in its healthcare; the enablers have positive influence. Due to this the current study considers the following hypotheses regarding the enablers: Hypothesis 2e (H2e). Employee related enablers have positive influence on the adoption of KM in healthcare.

Knowledge Management Drivers of Healthcare
There are several studies regarding drivers of KM. Davenport, et al. [100] identified several drivers to KM; improved knowledge access, enhanced knowledge environment, management of knowledge as an asset, improved economic performance, improved knowledge transfer, improved service, improved decision making process, bringing innovation, creating job opportunities, and decentralized decision making. Du Plessis [16] stated that improved quality of knowledge, knowledge hoarding, increase in efficiency, organizational communication, efficient transfer of knowledge, improved IT, effective decision making, creating competitive advantage, reduction in knowledge loss, and treatment of knowledge as a commodity as the main drivers of KM. Darko, et al. [101] addressed the main drivers of KM as setting a culture for best practice adoption, making a standard for other organizations to follow, improved employee efficiency, job creation, improved well-being of customers, reduced costs, and improved image of the organization. Lee, et al. [102] only addressed two drivers of KM; improved knowledge quality and quick transfer of knowledge. Yu [103] identified the drivers of KM as reduced administrative cost, reduced service cost, and quick decision making.
The knowledge management drivers of healthcare are given in Table 3. Less costly service [100,103,118] Drivers also make it easy for the stakeholders to adopt KM in its healthcare like enablers; drivers have positive influence. Due to this the current study considers the following hypotheses regarding the drivers:

Research Methodology
The aim of this study is to identify and analyze the barriers, enablers and drivers (variables) of KM adoption in healthcare. This study has been divided into two-steps; first the variables were identified by conducting a comprehensive literature review by reviewing several peer reviewed journals. After identifying the variables, the fuzzy Delphi method (FDM) was utilized to narrow down to the most relevant variables. In the second step SEM is applied. The step-by-step methodology is given in Figure 1. Step-by-step methadology.

Fuzzy Delphi Method (FDM)
The Delphi method (DM) was initially used in the 1950s by the RAND corporation in their studies [119], at that time it was considered as a reliable technique since it considered the collective opinion of experts [120]. Despite the multiple disadvantages attached with DM e.g., expensive to execute, experts would seldom agree to an opinion, and the researcher had the facility to adjust the opinion to their benefit, DM is still a widely used method in KM studies. To overcome the flaws of the DM, the fuzzy theory was implemented [121][122][123]. The DM was improved further by Hsu and Yang [124], by applying the triangular fuzzy number to encompass the opinion of the expert, providing foundation to the FDM. In the triangular fuzzy numbers (TFNs) the maximum and minimum values of the expert opinion are taken into consideration, based on them the geometric mean is calculated to avoid statistical biasedness based on extreme values. This helps in the correct selection of variables, it is a simple method and gives proper weightage to the expert's opinion in the Step-by-step methadology.

Fuzzy Delphi Method (FDM)
The Delphi method (DM) was initially used in the 1950s by the RAND corporation in their studies [119], at that time it was considered as a reliable technique since it considered the collective opinion of experts [120]. Despite the multiple disadvantages attached with DM e.g., expensive to execute, experts would seldom agree to an opinion, and the researcher had the facility to adjust the opinion to their benefit, DM is still a widely used method in KM studies. To overcome the flaws of the DM, the fuzzy theory was implemented [121][122][123]. The DM was improved further by Hsu and Yang [124], by applying the triangular fuzzy number to encompass the opinion of the expert, providing foundation to the FDM. In the triangular fuzzy numbers (TFNs) the maximum and minimum values of the expert opinion are taken into consideration, based on them the geometric mean is calculated to avoid statistical biasedness based on extreme values. This helps in the correct selection of variables, it is a simple method and gives proper weightage to the expert's opinion in the selection process [119]. The FDM was composed of two different rounds. At the end of first round, a facilitator prepared a summary which could help the experts for further screening (deletion or addition) the barriers, enablers and drivers. The step-by-step approach to obtain results for FDM are as follows: (1) Distribute the questionnaire and obtain response and preference for each barrier, enabler or driver through TFNs. (2) At the second step, fuzzy weights W wk obtained through TFNs were transformed into one single value V k by utilizing the center of gravity technique: (Where V k is the threshold criteria for rejection or selection of the appropriate item, Min represent the minimum value of TFNs, GM shows the geometric mean, and Max represent the maximum value of TFNs). (3) After two rounds, facilitator adopted the questions according to the threshold criteria that were the part of final questionnaires distributed in respondents.

Data Collection
A questionnaire is a systematic method of data collection, it is said to be effective and helps in reaching the objective and is quantifiable. That is why questionnaires have been used in this study to collect data regarding barriers, enablers, and drivers effecting KM adoption in the healthcare of Pakistan. The population considered for the study consists of stakeholders that possess knowledge about KM adoption in the healthcare industry of Pakistan. The stakeholders included Federal and Provincial Ministry of Health employees, hospital administration, doctors, nurses, dentists, and patients in major cities (Peshawar, Mardan, Abbotabad, and Islamabad). Due to this, the nonprobability sampling technique has been adopted. When random sampling cannot be done then the researcher can select the participant based on the participant's willingness to take part in the research. Hence a combination of two techniques was used, convenience sampling and snowball sampling to increase the overall sample size. Convenience sampling gave the ease of selecting respondents nearby, while snowball sampling helped in collecting data through references and networking. These techniques have been used in some management studies [125].
The questionnaire was developed based on the literature review of previous studies, the barriers mentioned in Table 1, enablers in Table 2 and drivers in Table 3. The questionnaire consists of five sections. In the first section a brief detail about the research is given, the objectives of the research, and the contact details of the researcher in case of quarries. In Section 2, the details of the respondent are recorded such as age, gender, organization type, ownership of organization, profession, and experience. In Sections 3-5, the respondents were asked to rate the barriers, enablers and drivers to KM adoption in the healthcare of Pakistan in their respective sections. The respondents were asked to rate the variables using the five-point Likert scale, in which 1 = Strongly disagree, 2 = Disagree, 3 = Neutral, 4 = Agree, and 5 = Strongly agree. The five-point Likert scale is used by many experts in research and it gives unambiguous results [126]. The initial questionnaire that was prepared was in English, but considering the national language of Pakistan, an Urdu version was also prepared. The translated version was prepared by two language experts.
The data collected for this study were obtained through self-collection, the research team, and short seminars. The team consisted of five individuals; two masters students and three masters graduates. They were briefed about the current research, its objectives and aims, to give them a good understanding about the study. These research team members were asked to collect data based on their contacts. The short fifteen-minute seminars were held in the public and private hospitals, attended by people willing to participate in the research. They were giving brief information about the current study and KM. After the seminar, a brief question and answer session was held to remove any queries and confusion, and after this, the participants were requested to fill out the questionnaire.
The sample size considered sufficient for SEM is 100 to 200 [127]. The response rate of healthcare is very low, and the experts consider above 42% as acceptable [128,129]. About 500 questionnaires were circulated among the stakeholders, out of which 255 were received. Of those, 18 of the questionnaires were removed because they gave invalid answers, 13 were removed because they were incomplete, resulting in 224 valid questionnaires making the response rate for the current study at 45%. The low response rate shows that KM adoption in healthcare is at the very initial stages and needs considerable attention of relevant authorities.
The demographic of the respondents is given in Table 4. The majority of respondents come in the age bracket of 31-40 which is 36.2%, with 37% respondents below this age and 26.8% above it. There were more male respondents (57.1%) as compared to females (42.9%). Of the respondents, 26.3% were working for the government in which 12.1 % were in government owned hospitals, while 14.3% were working in teaching hospitals. There were 38.8% of respondents working in privately owned organizations out of which, 10.7% were working in hospitals, 13.8% were working in teaching hospitals, 8% in medical centers, and 6.3% in the pharmaceutical companies. However, 34.8% of the respondents were not considered for this category since they were government employees or patients. The respondents had various different occupations 8.5% worked in the federal ministry of health, 12.9% in the provincial ministry of health, 11.2% were doctors, 8.9% dentists, 12.9% nurses, 15.6% administration, 13.4% patients, and 16.5% were technicians. Most of the respondents came in the experience bracket of 6-9 which is 32.6%, 29% of the respondents had less experience than this, and 25% had more experience than this.

Structural Equation Modeling (SEM)
SEM is a multivariate statistical tool that has been used several times for different types of research such as social science, applied science, health science etc. [130,131]. The SEM considers two types of variables; the observable, which can be easily measured, and latent, which are inferred from the observable. It is a technique which gives the researcher the ability to check the relationship between the observable variables and latent variables [132]. There are generally two approaches to using SEM., one is co-variance based (CB-SEM), and the other is partial least squares based (PLS-SEM). In this study the PLS-SEM was used because it is preferred by management researches [133,134], and it can be used with a small sample size [134]. In this study we have used a two-step method; first the measurement model was checked and then the structural model was checked, as proposed by Anderson and Gerbing [135]. In this study Statistical Package for Social Sciences (SPSS version 21) and AMOS (version 21) were used for analysis.
First of all, the goodness-of-fit measures must be checked, and to check it, several tests were performed. The chi-square (χ 2 ) test was done to check the relationships between the measured variables, and the degree of freedom (df) was also calculated referring to values that are free to vary in the final calculation. As suggested by Jöreskog and Sörbom [136] we used χ 2 /df as a measurement test. Other tests used for goodness-of-fit suggested by Hu and Bentler [137] were, (1) standardized root mean square (SRMR) used to calculate the difference between observed correlation and predicted correlation, (2) goodness-of-fit index (GFI) to determine how well the model fits the observation, (3) adjusted goodness-of-fit index (AGFI) was used to correct the GFI, to avoid being affected by latent variables, (4) normed fit index (NFI) is an incremental measure used to measure the goodness-of-fit, to insure the model was not affected by the number of variables, (5) comparative fit index (CFI) to check if the tested model was better than the alternate model, (6) root mean square error of approximation (RMSEA) to check the discrepancies in the hypothesized model, with parameters, and population covariance matrix by avoiding the issues of sample size.
The measurement model was tested using the (1) confirmatory factor analysis (CFA); it helped in testing the relationships between the variables, (2) Cronbach's alpha was used for internal consistency, (3) composite reliability (CR) was used to determine how much a latent variable effects the measurement of a measured variable, and (4) average variance extracted AVE was used to measure of the amount of variance that was captured by constructs in relation to the amount of variance due to measurement error.
Finally, after the measurement model was validated, the structural model was checked using the same goodness-of-fit measures. When the structural model is considered valid then the path coefficient, t-value and p-value will be used to check the hypothesis.

Validation and Reliability of Measurement Model
The model consisted of five latent variables, and eighteen indicators. The measurement model was tested using the confirmatory factor analysis (CFA) suggested by Hair, et al. [132]. After testing with CFA barrier OB4, SB2 and IB6 were removed for having a value lower than 0.5. After removing the barriers, the test was performed again, until a valid and reliable model was attained. The goodness-of-fit of measurement model is shown in Table 5, and the results show that the measurement model is valid. The χ 2 (70) has been calculated as well as the degree of freedom (32). Since χ 2 has some discrepancies [138], the test suggested by Jöreskog and Sörbom [136] was used, χ 2 /df = 2.180 which is within the limit ≤3 [138]. Other goodness-of-fit tests suggested by Hu and Bentler [137] were also calculated, and they all are given in Table 5. To check if the results of the latent variables are valid and reliable, the Cronbach's alpha, composite reliability and average variance extracted (AVE) were calculated. Since the values of Cronbach's alpha in Table 6 are between 0.856 and 0.715, which is more than 0.7, the data is reliable. Composite reliability is within 0.856 and 0.736, indicating that there is internal consistency in the measurement model since all values are greater than 0.6 [139]. The AVE values are between 0.605 and 0.527 which is more than 0.5. This indicates that more than 50% of the latent variables explain the variance in measurement items.   Table 7 shows the correlation of the latent variables and, since the values are less than the square root of their AVE, proves their validity. Table 8 shows the factor loading of each indicator, and as the values of the respective indicators are higher than others, the indicators have been correctly grouped. The structural equation model of barriers and knowledge management (KM) adoption derived from these calculations is given in Figure 2.

The Structural Model Validation and Reliability
After checking the measurement model, the structural model was tested. The structural model had five exogenous variables; organizational barriers, strategic barriers, technology barriers, resource barriers and individual barriers, and one endogenous variable which was knowledge management adoption in healthcare. The exogenous further had their indicators, three, two, two, three, and five respectively.
The structure model was validated using the same tests as the measurement model. The χ² has been calculated to be 76.51 and the degree of freedom as 40, hence χ²/df = 1.913. The other test results are also given in Table 9. To get the results displayed in Table 9, the bootstrapping technique was used. The path coefficient, t-value and p-value were calculated. The path coefficient shows the influence of independent variables on the dependent variables [23]. If the value of the path coefficient is between 0.1 and 0.3 the influence is weak, between 0.3 and 0.5 the influence is moderate, between 0.5 and 1

The Structural Model Validation and Reliability
After checking the measurement model, the structural model was tested. The structural model had five exogenous variables; organizational barriers, strategic barriers, technology barriers, resource barriers and individual barriers, and one endogenous variable which was knowledge management adoption in healthcare. The exogenous further had their indicators, three, two, two, three, and five respectively.
The structure model was validated using the same tests as the measurement model. The χ 2 has been calculated to be 76.51 and the degree of freedom as 40, hence χ 2 /df = 1.913. The other test results are also given in Table 9. Table 9. Checking the goodness-of-fit for structural model of barriers.

Goodness-of-Fit Recommended Value * Result
The To get the results displayed in Table 10, the bootstrapping technique was used. The path coefficient, t-value and p-value were calculated. The path coefficient shows the influence of independent variables on the dependent variables [23]. If the value of the path coefficient is between 0.1 and 0.3 the influence is weak, between 0.3 and 0.5 the influence is moderate, between 0.5 and 1 the influence is strong. If the t-values are less than 1.65, 1.96 or 2.58, respectively, they are insignificant. Note: OB stands for organizational barriers, SB stands for Strategic barriers, TB stands for technology barriers, RB stands for resource barriers, IB stands for individual barriers, and KMAHC stands for knowledge management adoption in healthcare. *** shows that the path coefficient is significant at p < 0.01, ** shows that the path coefficient is significant at p < 0.05, and * shows that the path coefficient is significant at p < 0.10.
The results indicate that organizational barriers and strategic barriers both had a path coefficient of more than 0.5 and t-value more than 2.58, organizational barriers are statistically significant at 1% and strategic barriers at 10% respectively. Due to this, hypothesis H1a and H1b were both supported. The other hypothesis H1c, H1d, and H1e were not supported because the path coefficient and t-values were less than 1.65, 1.96 or 2.58, so they are insignificant. The results indicate that technology barriers, resource barriers and individual barriers have a relatively lesser impact on KM adoption in healthcare. The SEM is given in Figure 2. The R 2 , also called the coefficient of determination, was calculated to be 0.386, indicating the accuracy of the model.

Validation and Reliability of the Measurement Model
The CFA test was run on the enabler indicators of KM adoption, and the results showed that the factor loading of MRE4 and ITRE4 were less than 0.5, hence they were removed. After the removal of these indicators, the test was rerun until a valid model was developed. The χ 2 = 75 and the df was 38, making χ 2 /df = 1.974 which is less than 3, hence the model was considered reliable. Several other tests have also been applied and they are given in Table 11. Table 11. Checking the goodness-of-fit for measurement model of enablers.

Goodness-of-Fit Recommended Value * Result
The The Cronbach's alpha, composite reliability and average variance extracted (AVE) were calculated. The Cronbach's alpha value was greater than 0.7, between 0.803 and 0.752, indicating that the data was reliable. The composite reliability shows that there is internal consistency in the measurement model when all values are greater than 0.6 [139], and since the values of our result were between 0.880 and 0.729, it indicates that there is internal consistency. The AVE values were between 0.713 and 0.516 which is higher than the recommended 0.50, indicating that half of the variances have been explained by the indicators. This showed that the data was strong, reliable and valid. The details of the values are given in Table 12, validity of constructs for enablers have been checked in Table 13 and their cross loadings are given in Table 14. From these values the structural equation model of enablers and KM adoption is given in Figure 3.

The structural Model Validation and Reliability
The structural model was tested after the measurement model. The structural model had five exogenous variables, management related enablers with five indicators, government related enablers with two indicators, information technology related enablers with three indicators,

The structural Model Validation and Reliability
The structural model was tested after the measurement model. The structural model had five exogenous variables, management related enablers with five indicators, government related enablers with two indicators, information technology related enablers with three indicators, customer related enabler with one indicator, employee related enablers with four indicators, and one endogenous variable knowledge management adoption in healthcare.
Same tests were made on structural model validation as on the measurement model. The χ 2 has been calculated to be 79 and the degree of freedom as 43, hence χ 2 /df = 1.837. The other test results are given in Table 15.
To get the results displayed in Table 16 the bootstrapping technique was used. The path coefficient, t-value and p-value were calculated. The path coefficient showed the influence of independent variables on the dependent variables [23]. If the value of the path coefficient is between 0.1 and 0.3 the influence is weak, between 0.3 and 0.5 the influence is moderate, and between 0.5 and 1 the influence is strong. If the t-values are less than 1.65, 1.96 or 2.58, respectively, they are insignificant.  Note: MRE stands for management related enablers, GRE stands for government related enablers, ITRE stands for information technology related enablers, CRE stands for customer related enablers, ERE stands for employee related enablers, and KMAHC stands for knowledge management adoption in healthcare. *** shows that the path coefficient is significant at p < 0.01, ** shows that the path coefficient is significant at p < 0.05, and * shows that the path coefficient is significant at p < 0.10.
Since the government related enablers had a path coefficient of more than 0.5, a t-value of more than 2.58, and was statistically significant at 1%, the hypothesis H2b has been supported. The other hypothesis H2a, H2c, H2d, and H2e were not supported because the path coefficient and t-values were less. The results indicate that management related enablers, information technology related enablers, customer related enabler, and employee related enablers have a relatively less impact on KM adoption in healthcare. The SEM is given in Figure 3. The R 2 , also called the coefficient of determination, was calculated to be 0.526, indicating the accuracy of the model.

Validation and Reliability of the Measurement Model
After doing the CFA test, HCRD4 and PRD1 were removed because their factor loading was less than 0.50. The test was continuously redone untill a valid model was achieved. The results of goodness-of-fit tests are given in Table 17. The Cronbach's alpha, composite reliability, and average variance extracted (AVE) were all within approved limits given in Table 18. The validity of constructs for drivers have been checked in Table 19 and their cross loadings are given in Table 20. From these values the structural equation model of drivers and KM adoption is given in Figure 4. Table 17. Checking the goodness-of-fit for measurement model of drivers.

Goodness-of-Fit Recommended Value * Result
The Chi Square (χ 2 ) N/A 132.83 degree of freedom (df) N/A 73

The Structural Model Validation and Reliability
The structural model had five exogenous variables, healthcare related drivers with four indicators, performance-based drivers with three indicators, communication related drivers with three indicators, knowledge related drivers with four indicators, patient related drivers with four indicators, and one endogenous variable knowledge management adoption in healthcare. The results of the goodness-of-fit are given in Table 21.
The bootstrapping technique was used to get the results given in Table 22. The path coefficient, t-value and p-value were calculated. The results indicated that healthcare related drivers and performance-based drivers both have a path coefficient of more than 0.5 and t-value more than 2.58, that healthcare related drivers are statistically significant at 1%, and performance-based drivers at 5% respectively. Due to this, hypothesis H3a and H3b were both supported. The other hypothesis H3c, H3d, and H3e were not supported because the path coefficient and t-values were less. The results indicate that communication related drivers, knowledge related drivers, and patient related drivers have less impact on KM adoption in healthcare. The SEM is given in Figure 4. The R 2 , also called the coefficient of determination, was calculated to be 0.485, indicating the accuracy of the model.  Note: HCRD stands for healthcare related drivers, PBD stands for performance-based drivers, CRD stands for communication related drivers, KRD stands for knowledge related drivers, PRD stands for patient related drivers, and KMAHC stands for knowledge management adoption in healthcare. *** shows that the path coefficient is significant at p < 0.01, ** shows that the path coefficient is significant at p < 0.05, and * shows that the path coefficient is significant at p < 0.10

Barriers
Barriers are variables that negatively influence the adoption of KM in the healthcare. Among the selected barriers, organizational barriers and strategic barriers hinder the adoption of KM the most. In an organization, the top management plays a vital role. If top management of the healthcare does not support the implementation of KM in healthcare, then it is one of the most critical barriers [45][46][47][48][49][50][51]. The top management of healthcare must give a clear vision and create an atmosphere where knowledge sharing is encouraged in order to ensure effective KM adoption in healthcare. Other than the top management, the structure and culture of healthcare must be considered as well. Organizational structure helps in task allocation, coordination, and supervision. It also controls the flow of information [140]. If the structure of the organization does not allow the flow of information then it will prove to be a barrier [2,13,45,[50][51][52][53][54]. Ichijo, et al. [12] pointed out that healthcare firms should maintain consistency between their structures to put their knowledge to use. Organization culture can be a critical problem when it comes to successful KM implementation [2,13,45,[50][51][52][53][54]. Culture is very important for the transferring of knowledge between employees [141]. A culture that encourages knowledge sharing is critical for KM success. Such a culture requires the healthcare employee to get together and exchange ideas; culture helps in collaboration and motivates the healthcare employees to work productively. Whenever a new plan is to be launched in an organization, its culture is considered carefully because the employees are involved. If KM is to be introduced in healthcare, the structure and culture must be taken into consideration. The infrastructure of Pakistan healthcare is very large. It is headed by the Ministry of Health (MoH); they should appoint heads of departments that possess the ability to make decisions on their own, rather than following the bureaucratic procedures. The top management of each department must create a structure that encourages the flow of information and create a learning environment. This would make KM adoption easier.
Strategic planning is also very important for the execution of KM. Ineffective strategic planning will prove to be a barrier [51,55,56]. Without effective strategic planning, it will be impossible to achieve KM [142] in healthcare. To implement KM in healthcare a clear strategy has to be made, one that everyone understands, and its goals, purpose, and objectives must be clear. Strategic planning is crucial for KM implementation in healthcare for sustainable competitive advantage and survival in the international market. Uncertainty about the effectiveness of KM may also prove to be a barrier. It is, however, not considered to be very critical, but it cannot be ignored [45,59]. To implement KM takes a long time, and it takes even longer to see the positive changes it gives. Due to this, the concept of KM according to the employees is not worth the effort/resources. The Government of Pakistan (GoP) has tried its best to improve healthcare by developing several strategies [143], but are currently not successful. The National Health Vision (NHV) [144] was approved in 2016 and is trying their level best to achieve it. The GoP is looking for new methods and is considering KM as an option.
Technology barriers, resource barriers and individual barriers also hinder the implementation of KM but it is not considered as significant in the case of Pakistan. The system of KM might be complex and difficult to implement or there might be difficulty in integrating it with the existing system. Resource barriers refer to cost of implementation and other resources needed for KM implementation. The individual barriers refer to conflicts, lack of motivation and resistance to change. Pakistan has recently invested heavily in getting new technology to improve administration and new machines for healthcare to improve patient service and has spent $3.04 billion [145]. It has increased its budget over the years to overcome other issues such as employee motivation and strikes [36].

Enablers
Enablers are variables that positively influence the adoption of KM in the healthcare. The results suggest that the government related enablers support the implementation the most. The government policies highly affect all organizations [89][90][91] in the public sector, and since the Pakistan healthcare sector has a large infrastructure it is highly influenced by policies [146]. The GoP provides basic healthcare through 5334 Basic Health Units (BHUs) and 560 Rural Health Centers (RHCs), secondary care through 919 Tehsil Headquarter Hospitals (THQs), District Headquarter Hospitals (DHQs) and it is estimated that there are about 96,430 private health establishments [146]. In Pakistan, the national polices and strategies are developed by the Federal MoH, which sets the goals and objectives. Whereas according to the constitution of Pakistan the provincial MoH is responsible for its deliverance and execution, except in the federally administrated areas. Favorable healthcare sector policies support the implementation of KM in healthcare. In the healthcare sector, clear long term strategic planning for implementation of knowledge management is also most critical for success [18,19,92]. Perera and Peiró [147] have stated that strategic planning is very important for all healthcare organizations, the short, medium, and long term vision and mission must be clear.
The GoP vision is "health for all". Pakistan, to improve its healthcare, has developed several policies over the years such as the National Health Policy (2001) [146], the Health Sector Vision (2005-2010) [146], and NHV (2016) [144]. The NHV was developed in 2010, it took a long time to be approved by both federal and provincial MoH. This document states the vision, mission, values and targets for 2025 of the healthcare very clearly so that Pakistan can improve its health standards. Pakistan has also signed international treaties such as the Millennium Development Goals (2000) setting targets for 2015 [148], and Sustainable Development Goals for healthcare setting targets for 2030 made by the United Nations (UN) [149]. The GoP has increased the budget of healthcare over the years to make sure it is not the shortage of money that is hindering its healthcare service [36]. Pakistan, realizing its weakness in the healthcare sector, is desperately trying to improve it and is now looking for new methods and is willing to adopt KM.
There are other enablers to healthcare as well such as management related enablers, information technology related enablers, customer related enablers, and employee related enablers, but they are considered as less supportive. If the management is supportive, there is a learning environment, and the employees trust each other, it will be easy to implement KM. The role of information technology cannot be avoided since it helps in storage and quick transferring of information. The customers help the process by giving constant feedback. If the employees are motivated, well trained and empowered they will perform better and help with implementation of KM in healthcare.

Drivers
Drivers are variables that positively influence the adoption of KM in the healthcare. The drivers that most significantly affect the adoption of KM are healthcare related drivers and performance-based drivers. Almost all the organizations have realized the importance of knowledge as an asset or commodity and are adopting KM to gain sustainable competitive advantage [16]. Globalization has increased the sense of competition among organizations, and healthcare is no exception. These days medical tourism has increased, patients are looking for places where they can get the best treatment [150]. This is why healthcare should be able to rapidly adjust to changes in the environment. If the healthcare sector is able to do this, it will certainly be setting the standard for other sectors by focusing on the best practices and utilizing the minimum of resources [151]. Over time other sectors will adopt the practices of healthcare [152]. However, the change in the health sector will not come overnight; it will take its time.
Pakistan progress in the healthcare has always been hindered by diseases and outbreaks [153]. Pakistan also suffers due to its large population, growth rate [35], and limited resources [143]. Nevertheless, Pakistan realizes the problems in its healthcare, and it has developed several polices and signed international treaties to improve performance. Pakistan is now considering KM because it understands the advantages that KM has to offer in the healthcare of Pakistan. It wants to bring a competitive advantage in its healthcare so that it can improve its reputation and get a share of the international market.
The performance is a critical factor of healthcare, and if there is effective decision making it will considerably improve the administration of healthcare. Decisions have to be made at many levels; top, middle and lower levels. The decisions made at the patient level are the most critical, as they have to be effective to reduce medical errors [154]. Pakistan healthcare generally has a centralized decision-making process. It will have to adopt a certain level of decentralization to ensure the quick flow of knowledge, and quick and effective decision making. This will also result in reduced utilization of resources by quickly dealing with the patients.
The other drivers, communication related drivers, knowledge related drivers and patient related drivers also support KM adoption but they are not as supportive. Communication is an important part of knowledge sharing. If there is communication between departments and other organizations, the knowledge will flow freely. Similarly, if there is a learning environment in the organization, the loss of knowledge will be less, and will create trust among the employees. This will create improvement in patient service resulting in less cost.

Conclusions
KM adoption has always been considered a source of sustainable competitive advantage. There are many barriers, enablers, and drivers that will influence its adoption. There have been very few studies in the area of KM in healthcare of developing countries [20,21]. This study was undertaken to check the quantitative influence of the variables on the adoption of KM in the healthcare of Pakistan. This study employs the SEM technique for the analysis of the variables. The data was collected via questionnaires, by five research representatives with knowledge of KM and several short seminars. The result of the study shows that organizational barriers and strategic barriers have a negative influence (barriers), whereas government related enablers (enablers), healthcare related drivers, and performance-based drivers (drivers) have a positive influence on KM adoption.
The reason of this study is to give a clear idea of KM adoption in the healthcare of Pakistan. The findings of this research will help the relevant authorities of Pakistan (government, hospitals, unions, staff, and etc.) get a better understanding of the barriers, enablers and drivers. The results show that the barriers can be overcome by the enablers and drivers. The organizational and strategic barriers are the main barriers. They need to be addressed in a way that reduces their influence. It can be done by developing suitable government policies (enabler) that encourage the flow of knowledge and make it easier to implement KM. Similarly, the adoption of KM will give the healthcare organizations a sustainable competitive advantage and improve their performance by effective decision making. This in return makes the healthcare sector a benchmark for other sectors and developing countries.
In this study there are a few limitations; the sample size and coverage were sufficient for the current study to apply SEM but they can be increased in future. The study, however, gives a good idea of how barriers, enablers and drivers influence KM adoption in healthcare. This study can be considered for other developing countries but it is more relevant to the situation of Pakistan.
In the future this study can be conducted again because barriers, enablers and drivers change with the passage of time depending on the phase of implementation. These studies might give further insight into the situation of KM in the healthcare of Pakistan. The same style of study can be used by researchers to determine the barriers, enablers and drivers in their respective developing countries.