Introduction
The knowledge of statistics is essential for interpreting research in health sciences and understanding evidence-based medicine. With growing literature on evidence-based medicine and its increasing utilization in clinical practice, learning statistics has become vital for practicing physicians (Windish et al., Reference Windish, Huot and Green2007; Sami, Reference Sami2010). Although physicians increasingly realize the importance of statistics, various studies have shown physicians’ inadequacies in the knowledge of statistics and in the skills of applying statistical tests (West and Ficalora, Reference West and Ficalora2007; Kilic and Celik, Reference Kilic and Celik2013). Research has also shown that practicing physicians did not have a proper understanding of statistical tests commonly used in published research, and had difficulty in interpreting study results (Windish et al., Reference Windish, Huot and Green2007).
Primary health care (PHC) is the cornerstone of the health care system. Quality health services at primary care are vital for a healthy community. For quality health care, the primary care physicians must keep abreast of current clinical information to practice evidence-based medicine. To understand evidence-based medicine literature, primary care physicians require to comprehend the research studies and should be able to interpret their results. It has been reported that the physicians perceive statistics as a difficult subject, and this perception affects their attitudes to statistics (Kilic and Celik, Reference Kilic and Celik2013).
Studies have shown that attitudes play a significant role in learning statistics (Zhang et al., Reference Zhang, Shang, Wang, Zhao, Li, Xu and Su2012). Thus, it is important to determine the attitudes of the target participants, before planning educational programs related to statistics. Various studies have been conducted to determine the attitudes of students to statistics in different fields including psychology and nursing (Zimprich, Reference Zimprich2012; Baghi and Kornides, Reference Baghi and Kornides2013; Mathew and Aktan, Reference Mathew and Aktan2014). Some studies were also conducted to assess physicians’ attitudes to statistics (Ercan et al., Reference Ercan, Ozkaya, Ocakoglu, Yazici, Sezer, Ediz and Kan2008; Rashid and Subramaniam, Reference Rashid and Subramaniam2012; Zhang et al., Reference Zhang, Shang, Wang, Zhao, Li, Xu and Su2012; Kilic and Celik, Reference Kilic and Celik2013). However, there is a dearth of literature regarding attitudes to statistics among physicians in general and PHC physicians in particular. On literature search, no such study could be found from Saudi Arabia. The Ministry of Health (MOH) is the main provider of PHC in Saudi Arabia. Most PHC doctors are general physicians with a degree of bachelors in medicine. These physicians are professional leaders at PHC centers. The MOH has a variety of PHC programs including Antenatal Care and Childhood Immunization. Data for these programs are collected at the level of PHC centers. Physicians are expected to comprehend those data, and support the allied staff in data management at the PHC centers. Thus, strengthening physicians’ skills in statistics is required to carry out the activities such as assessment of program performance, quality assurance, and performance improvement projects. In Saudi Arabia, there is a culturally diverse health care workforce with varied educational and professional experiences with regards to statistics. In this situation, it is even more important to determine the physicians’ attitudes to statistics. Understanding physicians’ attitudes to statistics will be helpful in planning in-service training related to statistics.
To understand the attitudes to statistics, held by PHC physicians, we surveyed all PHC physicians in Qassim province to measure these attitudes and explore related factors. The purpose of the study was to investigate PHC physicians’ attitudes to statistics; to determine PHC physicians’ self-reported knowledge level in statistics; and to determine PHC physicians’ perceived training needs in statistics. The results of this study will be helpful in designing a training course on statistics for PHC physicians.
Theoretical framework
The study was based on theory of planned behavior (Ajzen, Reference Ajzen1991). According to this theory, human behavior is logical and there is a reason for people behaving the way that they do (DeBarr, Reference DeBarr2004). The basic concept of the theory is that behavior change occurs if the person has an intention to change, and two main factors influence the intentions; attitudes toward the behavior and subjective norm toward the behavior (DeBarr, Reference DeBarr2004).
As human behavior is logical, one should be able to explain or predict it. Behavioral intentions are the best predictors of human behavior, and the best predictors of a person’s stated behavioral intentions are the attitude and subjective norm. Attitudes are the positive or negative feelings toward a concept, behavior, or an individual. The attitude toward a particular behavior can be predicted by examining what is the expectation regarding the outcome of a particular action and the likelihood of achieving this outcome (DeBarr, Reference DeBarr2004). Subjective norms are a positive or negative judgment of the friends or colleagues about a particular behavior. Subjective norms are affected by normative beliefs, which means that a person is concerned about the approval of adopted behavior by the significant others, and is motivated to comply with their potential approval (Schiavo, Reference Schiavo2007). This theory emphasizes that actual behavioral control is dependent upon resources and opportunity (DeBarr, Reference DeBarr2004). Perceived control over the resources and opportunities required to perform a behavior, is believed to be an important aspect of behavior change processes. Theory of planned behavior has been applied to various areas such as education and health care.
Methods
The study was a cross-sectional survey. The survey included physicians working in PHC centers in Qassim province. A semi-structured, self-administered questionnaire was used as the survey instrument. The primary purpose of the questionnaire was to determine the physicians’ attitudes to statistics.
The concepts of theory of planned behavior were applied in designing this survey. According to the theory of planned behavior, the intentions can be assessed by attitudes toward the desired behavior and subjective norms toward the behavior (DeBarr, Reference DeBarr2004). The statements of the questionnaire were developed to determine these components of theory of planned behavior. Thus, the survey also helped in determining the intention of the participants to adopt the behavior of using statistics in their routine daily work.
The questionnaire comprised questions for gathering information about (a) baseline profile of the participants concerning their demographic information as well as data about statistics courses attendance; (b) attitudes to statistics; (c) participants’ opinion concerning perceived need of training in statistics. The attitude instrument was developed through literature consultation, and the statements were adapted from validated existing surveys that addressed attitudes to statistics (West and Ficalora, Reference West and Ficalora2007; Windish et al., Reference Windish, Huot and Green2007; Zhang et al., Reference Zhang, Shang, Wang, Zhao, Li, Xu and Su2012; Zimprich, Reference Zimprich2012). In our study, the attitude was determined by a 20-item survey. The 20-item survey consisted of four subscales: (i) general perceptions: assessed participants’ positive and negative feelings about statistics (eg, ‘I find statistics to be very logical and clear’); (ii) perceptions of knowledge and training: described participants’ perceptions about how well they are trained in statistics (eg, ‘My training in statistics is adequate for my needs’); (iii) perceptions of statistics and evidence-based medicine: captured attitudes about the usefulness and importance of statistics and evidence-based medicine (eg, ‘Statistics is an important part of evidence-based medicine’); and (iv) perceptions of future learning: measured participants’ opinions about learning statistics in future (eg, ‘I would like to study statistics’). Each subscale contained five statements pertaining to the topic of the subscale. The items were measured on a five-point Likert scale, with ‘1’ indicating that the participants disagreed strongly with the statement and ‘5’ suggesting that they agreed strongly with the statement. Seven of the 20 items were negatively worded; the remaining 13 items were positively worded. During analysis, negatively worded items were reversed so that higher scores always indicated a more positive attitude.
The questionnaire was piloted with the physicians working in health care facilities in Qassim. The physicians were asked to provide comments about the questionnaire, including length of the questionnaire, clarity of questions, and response options. The necessary changes were made according to their comments.
In Saudi Arabia, well-established PHC system exists. The MOH operates 2259 PHC centers in Saudi Arabia. Each center provides preventive, curative, and health education activities (Ministry of Health, 2013). The managerial structure of PHC in health directorates of Saudi provinces comprises mid-level managers in each health sector or supervisory office. A health sector or supervisory office is an administrative unit that supervises and manages health services in a specified number of PHC centers. Qassim is the northern central province of Saudi Arabia with a population of >1.3 million, during the year 2013. Saudi nationals comprise 1.01 million, whereas non-Saudis comprise 0.33 million of the population (Ministry of Health, 2013). In 2013, there were 159 PHC centers in Qassim (Ministry of Health, 2013). These centers primarily serve the Saudi nationals and also provide health care facilities to a limited number of non-Saudis.
All PHC physicians in Qassim province were invited to participate. A total of 416 eligible study participants were invited by an official letter issued by the highest authority for Public Health Administration, Qassim. The survey was conducted from August 14 to October 30, 2014. The survey instrument was distributed through an official mailing system. A package in an envelope containing a cover letter, informed consent forms, and the questionnaires, was posted for physicians working in the respective health sector, through the official mail to all supervisory offices in Qassim province. A return date of two weeks after receipt of the questionnaire was requested. Non-responders were followed up by telephone reminders. The district supervisors returned the completed questionnaires via official mail.
Data were analyzed using the software IBM Statistical Package for Social Sciences for Windows version 21 (Chicago, IL, 2012). The negative statements were changed to positive, with a score of one being counted as five and a score of five as one. Missing data on single statements were replaced by mean value substitution when processing the subscale (Green and Salkind, Reference Green and Salkind2011). The mean value substitution was only done if out of the total five, two or less values were missing. Descriptive statistics were computed such as mean and standard deviation, frequency and proportions. Likert-type items were compared using Pearson’s correlation coefficient, paired t-test, and one-way analysis of variance (ANOVA) F-test. A P-value of 0.05 or less was considered as statistically significant. Cronbach’s αreliability coefficient of the instrument was assessed.
Content analysis of the open-ended question about suggestions for topics of the statistics training was conducted. The responses were grouped into categories emergent from the data. These categories comprised single words or phrases that described and summarized the responses given. Any participant’s response could appear under more than one category if mentioning several topics relevant to different categories; thus, there were more comments than participants responding to the question. Differences in ideas about coding, including naming and constitution of categories, and classification of responses were discussed and finalized after developing consensus among the authors.
Ethical approval was obtained from Regional Research Ethics Committee. An informed consent was taken from the participants. The questionnaires were kept anonymous. Each questionnaire was given a unique identifier that was used in the management of data.
Results
Of 416 eligible participants, 338 PHC physicians working in Public Health Administration, Qassim, participated in the study. Thus, the survey response rate was 81.25% (338/416). The mean age of the participants was 40.8 (±9.3) years, whereas the mean duration since graduation was 14.8 (±8.9) with a minimum of less than one year and a maximum of 38 years. The demographic data of the participants are shown in Table 1. The participants included 204 (60.9%) males and 131 (39.1%) females. The frequencies of nationalities showed that 43.0% of the participants were Sudanese, followed by Egyptians (33.6%) and Pakistanis (7.9%). A majority (84.8%) of the participants were graduates having only a bachelors degree in medicine.
Table 1 Demographic characteristics of the participants (n=338)
The participants were asked questions to determine their interest and training in statistics (Table 2). Over half the participants (59.6%) reported being trained in statistics during academic training and 15.2% attended statistics course after graduation. A majority of the participants (95.8%) had not attended any statistics course during last five years. Only a quarter (25.2%) of the participants reported having looked/read a statistical report from Saudi Arabia and Qassim during past six months.
Table 2 Participants’ training and interest in statistics
The participants’ self-reported knowledge of statistics was determined on a scale of 1–10. The median and mode of participants’ self-reported knowledge was five, with a minimum of 1 and a maximum of 10. The majority (73.6%) of the participants self-assessed their level of statistics knowledge as five or below.
Attitudes to statistics
The Cronbach’s α was calculated using responses to all 20 items on attitude scale. It was 0.76, showing sufficient internal consistency. The total score for attitudes to statistics scale could be between 20 and 100. Higher scores showed a positive attitude to statistics. Most of the participants showed a high total score on the attitudes scale. The total mean score on attitudes to statistics was 71.14 (±7.73) with a minimum of 45 and a maximum of 91. The mean total attitude score among males was 72.35 (±7.52), whereas for females it was 69.34 (±7.73). This difference was statistically significant at P<0.001.
Table 3 shows the mean scores of the participants on the four subscales. The participants responded to the ‘perceptions of statistics and evidence-based medicine’ subscale with highest scores. Having the total value between 5 and 25, the mean score for this subscale was 21.54 (±2.71) with a minimum of 12 and a maximum of 25. The participants responded in the negative manner to ‘perceptions of knowledge and training’ subscale, as they had the lowest score on this factor. The mean score for knowledge and training perception subscale was 14.08 (±3.08) with a minimum of five and a maximum of 23.
Table 3 Mean subscale scores for participants’ attitudes to statistics
Total score for each item=25.
Table 4 displays the mean scores of the individual items. The item to which the participants responded with the highest score, was ‘evidence-based medicine is important for clinical practice’ (4.53±0.70), followed by ‘statistics is an important part of evidence-based medicine’ (4.51±0.78). The item to which the participants responded to with the lowest score (2.14±0.96) was ‘learning statistics requires a great deal of efforts’ (Table 4).
Table 4 Mean statistics attitude scores for individual items
a The negative statements were recoded; higher scores mean positive attitudes.
Table 5 displays association of various factors with overall attitudes to statistics scores. An independent sample t-test was conducted to evaluate the hypothesis that males have a more positive attitude to statistics as compared with females. The test was significant, t (319)=3.464, P=0.001. Mean difference in scores among males and females was 3.01 and the 95% confidence interval for the difference in means ranged from 1.299 to 4.714.
Table 5 Association of various factors with overall statistics attitude scores
A ANOVA was conducted to evaluate the relationship between the qualification and overall attitude scores. The independent variable, qualification, included three levels: MBBS, Diploma, and Master degree. The ANOVA was significant, F(2, 259)=3.223, P=0.041. Post hoc tests were conducted to evaluate pairwise differences among the means. There was a significant difference in the mean scores between the MBBS and Master, P=0.047, but no significant difference was found between MBBS and Diploma; and Diploma and Master. Studying statistics during academic training and after graduate studies were both significantly associated with overall statistics attitude scores.
A linear regression analysis was conducted to evaluate the prediction of the total attitude scores from the age of the physician. Age was significantly related to the overall attitude scores. The correlation between age and total attitude score was 0.189 (P<0.01). Accuracy in predicting the total attitude scores was mild. Approximately, 3.3% of the variance of the total attitude score was accounted for by its linear relationship with age. Linear regression analysis showed no significant prediction by the number of statistics courses attended during last five years on total attitude scores.
A multiple regression analysis was conducted to evaluate how well the general perception measures predicted total attitude scores. The predictors were five items in the general perception subscale, whereas the dependent variable was total attitude scores. The linear combination of general perception measures was significantly related to the total attitude scores, F(5, 272)=136.736, P<0.0001. The adjusted r 2 indicated ~71% of the variance of total attitude scores in the sample can be accounted for by the general perception items.
The participants were asked if they had used the data in their health centers for any academic or educational purpose. Out of 338 total responses, 190 (56.2%) reported having used data for educational purposes. Mostly the data for hypertension, diabetes mellitus, and immunization were used for this purpose.
The vast majority of the participants responded affirmatively to the need for learning statistics. On enquiring about their perceived need to learn statistics, 314 (92.9%) participants showed their desire to learn statistics. Responding to the suggested topics, 237 (70.1%) participants provided a total of 318 suggestions. Out of the total suggestions, 26.7% were related to epidemiology topics such as study designs and topics concerning evidence-based medicine. The rest of the suggestions included general statistics covering all topics of statistics; specific topics of descriptive statistics and inferential statistics. Topics unrelated to statistics, such as clinical features and management of chronic diseases, were proposed in 23.3% suggestions.
Discussion
The results of our study suggest that PHC physicians are aware of the importance of statistics in their professional career, and also realize their lack of competence in statistics. Other studies have also reported similar findings among physicians (West and Ficalora, Reference West and Ficalora2007). Our survey had a high response rate and good completeness of the questions, which may indicate the interest of study participants in statistics. A substantial proportion of our study participants (59.6%) had studied statistics during their academic training, which can be explained by the fact that nowadays statistics is a core module of many medical schools’ curricula (Dommisse and Joubert, Reference Dommisse and Joubert2009). This finding is in contrast to another study in which a survey of 62 physicians in surgical residency programs found that only 33% had formal statistics teaching in their medical schools (Cheatham, Reference Cheatham2000). This difference may be explained by the fact that the study by Cheatham (Reference Cheatham2000) included the respondents graduating before the year 2000. Anecdotal evidence suggests that since then the curricula for most of the medical institutions have been reviewed with the inclusion of statistics or equivalent courses.
In our study, only 15.2% participants had attended a statistics course after graduation, and the vast majority (95.8%) had not attended a statistics course during last five years. This finding may reflect the lack of availability of statistics training courses at PHC level. It underscores the importance of conducting training courses for PHC physicians to familiarize them with basic statistics. In spite of substantial number of relevant documents available online as well as in the form of hard copies, only a small proportion (25%) of participants had read/looked at reports about health statistics in Qassim or Saudi Arabia. This finding may reflect a lack of awareness of the PHC physicians about the availability of these reports. Moreover, they are heavily committed to their clinical duties, and are not obligated to conduct research or prepare scientific reports related to their work. Thus, the work environment is not very supportive in encouraging the physicians to engage in research, scientific, or academic activities. Lack of interest in statistical reports can also be explained by the self-reported low level of knowledge of statistics in our study, with almost three-quarter of the participants self-assessing themselves as below average. This finding is reported in other studies in which physicians have self-reported lack of competence in statistics (West and Ficalora, Reference West and Ficalora2007). Ercan et al. (Reference Ercan, Ozkaya, Ocakoglu, Yazici, Sezer, Ediz and Kan2008) also found an insufficient knowledge of statistics among physicians.
The results of this study indicate that the participating PHC physicians had a positive attitude to statistics. Several studies conducted with the health care workers, including physicians (West and Ficalora, Reference West and Ficalora2007), medical students (Hannigan et al., Reference Hannigan, Hegarty and McGrath2014; Stanisavljevic et al., Reference Stanisavljevic, Trajkovic, Marinkovic, Bukumiric, Cirkovic and Milic2014), and nurses (Baghi and Kornides, Reference Baghi and Kornides2013), have shown a positive attitude to statistics, and the participants appreciated the importance of statistics in their professional career.
In our study, significant associations were found between total attitude scores and age, qualification, and studying statistics during academic training and after graduation. Other studies have also found that the realization of the importance of statistics increases with increase in age, qualification, and status. Ercan et al. (Reference Ercan, Ozkaya, Ocakoglu, Yazici, Sezer, Ediz and Kan2008) found that the senior staff, involved in academic activities, believed more in the importance of statistics in professional carrier as compared with residents and students, in a medical school in Turkey. However, other studies from Malaysia (Daher and Amin, Reference Daher and Amin2010), Pakistan (Ahmad et al., Reference Ahmad, Zehra, Omair and Anjum2009), and Croatia (Hren et al., Reference Hren, Lukić, Marusić, Vodopivec, Vujaklija, Hrabak and Marusić2004) have shown positive attitude of medical students to statistics and its significance in medical curriculum.
In our study, males scored higher on attitude scale showing a more positive attitude in comparison to females. Although the difference between mean scores was small, it was statistically significant. Another study conducted among medical students found that on attitude scales, the males scored higher as compared with females; however, the difference was not statistically significant (Hannigan et al., Reference Hannigan, Hegarty and McGrath2014).
The analysis of the four subscales of attitudes to statistics scale showed interesting findings. The results of ‘perceptions of statistics and evidence-based medicine’ subscale analysis, showed highest scores, followed by ‘perceptions of future learning’. ‘Perceptions of knowledge and training’ was the least scoring subscale.
Overall, the participants had a positive general perception toward statistics. Among the general perception items, ‘Statistics should be a required part of my profession’s skills’ scored the highest (4.21±0.79) reflecting realization of importance of statistics in their professional life and a desire to have knowledge and skills of statistics.
‘Evidence-based medicine is important for clinical practice’, was the highest scoring (4.53±0.70) item of the subscale ‘perceptions of statistics and evidence-based medicine’. Supporting the finding of our study, previous studies have also shown the awareness of health care professionals regarding the importance of statistics in evidence-based medicine. A survey of practicing clinicians found that knowledge of statistics is considered useful in many aspects of clinical practice including comprehension of clinical guidelines and evidence summaries, explaining levels of risk to patients, and interpreting the results of a screening test (Swift et al., Reference Swift, Miles, Price, Shepstone and Leinster2009). Moreover, health care professionals require the knowledge of statistics for medical practice and understanding of medical issues (Freeman et al., Reference Freeman, Collier, Staniforth and Smith2008). The published medical literature cannot be critically evaluated without basic knowledge of statistics (Daher and Amin, Reference Daher and Amin2010; Stanisavljevic et al., Reference Stanisavljevic, Trajkovic, Marinkovic, Bukumiric, Cirkovic and Milic2014). It is also reported that the health care professionals who feel competent to understand research are more likely to practice evidence-based medicine (Baghi and Kornides, Reference Baghi and Kornides2013).
The scores on ‘perceptions of future learning’ subscale were the second highest. It shows that the physicians not only realize the importance of statistics and evidence-based medicine but are also keen to learn the subject of statistics. This is in accordance with the studies conducted with professionals in health care organizations where statistics and evidence-based medicine were considered important for their practice, and the need of proper understanding of statistics was clearly recognized by physicians (West and Ficalora, Reference West and Ficalora2007). The physicians believed that their routine work as a doctor required knowledge of epidemiology, statistics, and analysis skills (Miles et al., Reference Miles, Price, Swift, Shepstone and Leinster2010).
The participants scored the lowest in the ‘perceptions of knowledge and training’ subscale, indicating that they realize the inadequacy of their knowledge and training in statistics. Even physicians with statistical training and substantial research experience have been found to think that they lack required competence in statistics (West and Ficalora, Reference West and Ficalora2007). Studies have also demonstrated that practicing physicians had poor understanding of common statistical tests and their ability to interpret study results was limited (Windish et al., Reference Windish, Huot and Green2007). Factors related to the lack of knowledge and training include the perception of statistics as a difficult field, lack of sufficiently understanding the importance of statistics, inadequate statistics training in the medical schools, and lack of utilization of problem-based approach during statistics teaching (Bland, Reference Bland2004; Kilic and Celik, Reference Kilic and Celik2013). Moreover, a survey found that many physicians had not recognized the value of their undergraduate teaching in statistics, but had subsequently found the skills relevant to their career (Miles et al., Reference Miles, Price, Swift, Shepstone and Leinster2010).
In our study, the vast majority of participants (92.9%) expressed their perceived need to learn statistics. They suggested a number of topics for the training courses, which indicates their interest in statistics. However, many suggestions related to the topics of evidence-based medicine shows the clinical context of the PHC physicians.
Our study surveyed PHC physicians at a single province. Differences in qualification and work environment of participants, may limit generalizations to other settings. However, the participants in our study represented a diverse population of physicians from various Middle Eastern, South Asian, and African countries. Other provinces of the country have a similar composition of PHC physicians, making it possible to generalize our findings to other provinces. A self-administered questionnaire may also limit the validity of findings. Being a self-reporting survey, recall bias and misunderstanding of the questions may have occurred, contributing to inaccuracies in the data. However, efforts had been made to clear any ambiguity by making amendments after pretest. Non-response bias is minimal in this survey as the response rate was high, and mostly the non-responders were those who were on vacations, at the time of the survey.
Conclusion
Our study results suggest that the PHC physicians have a positive attitude to statistics. They realize their gaps in knowledge in statistics and are keen to improve their knowledge. However, surveys for determining attitudes need to be done periodically because of high turnover and diversity of the health care workforce in Saudi Arabia. Moreover, change in statistical skills and attitudes to statistics may occur because of changing curricula with a focus on biostatistics and research in medical schools.
Implications for policy and practice
To promote the use of statistics in PHC in Qassim, strategies must be adopted to involve PHC physicians in handling data, and to improve their statistics knowledge and skills. PHC physicians may be involved in managing data by making them realize that the available data are valuable in assessing their work. Regular statistics training programs and supportive work environment may be helpful in improving knowledge and skills of the physicians. A regular statistics course could be added in the annual Academic Calendar of Public Health Administration, developed according to the needs of PHC physicians. The statistics training should cover basic statistics and focus on the role of statistics in clinical decision making, and ability to interpret medical research. Building statistical skills using courses is expected to improve the skills especially related to the critical appraisal of the published literature. With an improved knowledge of statistics, better clinical care utilizing evidence-based medicine is expected, leading to an improvement in health care services.
Acknowledgments
The authors are grateful to all survey respondents for their valuable time and efforts in completing the survey. The authors also wish to thank all District Supervisors and Primary Health Care Centre Directors of Qassim province who actively participated in data collection. The authors are thankful to Ms Resa Bedeo, Office Secretary at Research and Information Unit, for her valuable secretarial tasks.
Financial Support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Conflicts of Interest
None.
Ethical Standards
The study was approved by Regional Research Ethics Committee, Qassim, Saudi Arabia.
