Study design and setting
The study was part of the EnPHC evaluation project conducted at 40 public primary care clinics. It is a quasi-experimental study with the primary objective of evaluating the impact of EnPHC interventions on processes of care and intermediate clinical outcomes among T2DM and hypertensive patients. The study collected information from three aspects: the clinics (facility survey), HCPs (provider survey) and patients (patient exit survey and retrospective chart review) across 20 intervention and control public clinics matched by the number of family medicine specialists and medical doctors, location of clinics and annual attendance. The effect of interventions on HCPs, which is a secondary objective of the evaluation project and the focus of this paper was evaluated using similar study design. Further details on the design and methods of the study have been described in a study protocol which is currently under journal review.
Study population
Our study included all HCPs who were directly involved in patient management (consisting of family medicine specialists, medical doctors, assistant medical officers (AMOs), nurses, pharmacists, dietitians, nutritionists, physiotherapists, occupational therapists and medical social workers). Medical doctors are licensed doctors with basic medical training whereas AMOs are similar to nurse practitioners in other countries. All HCPs who were present at the study clinics during the data collection period were invited to participate in the survey.
Data collection timeline
Baseline data collection was conducted between April and May 2017 where a self-administered provider questionnaire was distributed to each eligible HCP during the data collection period. EnPHC interventions were then implemented in July 2017 and were given three months to reach full capacity followed by a further 17 months of implementation (October 2017 to February 2019). A post-intervention survey was administered to the HCPs in the same clinics between March and April 2019.
Survey instrument
Provider questionnaire was developed based on the questions acquired from the General Practitioner (GP) questionnaire in Quality and Cost of Primary Care (QUALICOPC) study [27]. QUALICOPC is a multi-country study which evaluates quality, costs and equity of primary care system in Europe by using four sets of questionnaires. The QUALICOPC General Practitioner questionnaire was adapted and used for Malaysian QUALICOPC study which was conducted between 2015 and 2016. The results of the doctors’ job satisfaction in QUALICOPC Malaysia study have been published [28]. Modification and adaptation of the questions were done for our provider questionnaire and to accommodate for language proficiencies of other healthcare professionals as QUALICOPC General Practitioner questionnaire was initially designed only for doctors.
The English version of the questionnaire was translated to Malay language by two study collaborators who were fluent in both English and Malay. The Malay version of questionnaire was then back-translated into English by two independent translators according to World Health Organization recommendations [29]. Subsequently, the research team compared both versions of questionnaire and resolved any discrepancies to ensure that the translated items retained the same meaning as the original items. Eight HCPs were engaged from two public clinics, which were not part of the study sample, for the questionnaire pre-test. The provider questionnaire collected information on provider demographics and clinics characteristics, workload, quality of care from providers’ perspective, professional roles as well as job satisfaction.
Dependent variables
Job satisfaction was measured using the following six items: (i) “I feel that some parts of my work do not really make sense”, (ii) “My work still interests me as much as it ever did”, (iii) “My work is overloaded with unnecessary administrative detail”, (iv) “I have too much stress in my current job”, (v) “Being a healthcare provider is a well-respected job”, (vi) “In my work there is a good balance between effort and reward”. All statements were assessed on a 4-point Likert scale: 1 = strongly agree, 2 = agree, 3 = disagree and 4 = strongly disagree. Reverse coding was applied to responses for the following items to keep the scale in the same direction: “work still interesting”, “well-respected job”, “effort-reward balance”. The responses were coded in such a way that a higher score indicates higher job satisfaction (1 = low job satisfaction and 4 = high job satisfaction). For example, a high score on “well-respected job” reflects that the respondent strongly agreed with the statement, which indicates a high job satisfaction. The scores for each item were averaged and subsequently analysed per item.
Independent variables
Respondent characteristics and other details including age, gender, educational level, overall duration of practice in healthcare, duration of practice in primary care settings, hours spent per week on direct patient care and professional role were collected. The number of hours spent per week on direct patient care was used as a proxy for individual provider workload. The categories for professional role were: doctors (included family medicine specialists and medical doctors), AMOs, nurses and integrated specialized services personnel (pharmacists, dietitians, nutritionists, physiotherapists, occupational therapists as well as medical social workers). In addition, other clinic characteristics relevant for analysis such as location of the clinic (urban/rural) and study arm (intervention/control) were recorded.
Statistical Analysis
Categorical variables were reported in frequencies and percentages while continuous variables were presented as mean and standard deviation (SD). Sociodemographic characteristics of the intervention and control groups were compared using Chi-square and independent t tests.
To estimate the effects of EnPHC interventions on HCPs’ job satisfaction, difference-in-differences (DiD) analysis, which is a common method used in quasi-experimental studies was utilised. The impact of the intervention is estimated through the difference between two differences in the outcomes: (i) difference between the pre- and post-intervention periods within control group and (ii) difference between the pre- and post-intervention periods within intervention group [30,31]. In this study, repeated cross-sectional DiD analysis was conducted separately for each variable of job satisfaction using a multivariable linear regression model (equation 1) since there were new respondents during post-intervention phase. Hence, the change in job satisfaction following implementation of EnPHC interventions was estimated from DiD analysis after controlling for differences at baseline, HCP and clinic level covariates as the below equation:
All models used generalised estimating equations to adjust for clustering observations within clinics and to estimate change in mean score for job satisfaction using cluster robust standard error (SE). Multicollinearity of the covariates was checked and detected in one pair of covariates; age and overall duration of practice in healthcare. The variable age was deemed more relevant and hence kept in the analysis. A proportion of the HCPs was surveyed only once, at either baseline or the post-intervention phase. Sensitivity analyses were conducted among the subgroup of providers who responded at both baseline and post-intervention phases and these results were compared to the results in the main analysis.
Subgroup analyses were conducted to assess the changes in job satisfaction among all different professional roles using DiD analysis except for CCs which were only present in intervention groups. A non-parametric Wilcoxon Signed Rank test was used to analyse the impact of EnPHC interventions on job satisfaction among CCs. A p-value of <0.05 was considered as statistically significant. All statistical analyses were performed using “geepack” package in R version 3.5.3 in RStudio version 1.1.463 [32].