Evaluation of the reporting quality of observational studies in Master of Public Health’s dissertations in China: a systematic review

Backgrounds Master of public health (MPH) plays an important role in Chinese medical education, and the dissertations is an important part of MPH education. In MPH dissertations, most are observational studies. Compared with randomized controlled trial (RCT), observational studies are more prone to information bias. So, the reporting of the observational studies should be transparent and standard. But, no research on evaluating the reporting quality of the MPH dissertation has been found. Methods A systematic literature search was performed in the Wanfang database from January 1, 2014 to May 31, 2019. The Strengthening the Reporting of Observation Studies in Epidemiology (STROBE) statement was adopted to evaluate the reporting quality of the selected studies. Results The median of compliance with STROBE statement of 165 articles was 67.82%. The mean (standard deviation) of STROBE score was 14.3 (1.91). Five items/sub-items were 100% reported: background, objectives, study design, report numbers of individuals at each stage, and key result. Fifteen items/sub-items were reported by 75% or more. Reporting of methods and results was often omitted: missing data (12.73%), sensitivity analyses (3.03%), ow diagram (15.15%), and absolute risk (0%). Logistic regression analysis indicated that funding support (OR=13.98, 95% CI=4.37-44.70) and more published papers during postgraduate period (OR=2.77, 95% CI=1.02-7.54) were related to high reporting quality. We authors when studies.


Background
Public health in the 21st century faces problems that are very different from those in previous centuries [1]. With the development of economic in China, more and more public health problems are appearing, which poses a serious challenge to public health practitioners [2,3]. As of 2014, the total number of professional staff in China's public health institutions was only 87.5 million, which was well below the target of reaching 95 million in 2015. In addition, only 4.2% of public health professionals had a postgraduate degree [4]. The shortage of health professionals and highly educated health personnel has hindered the development of public health services in China. Therefore, it is urgent to optimize the training program and train more highly educated, application-oriented public health personnel. In order to ful ll this requirement, the Ministry of Education of China launched the full-time Master of Public Health (MPH) postgraduate program in 2009 [5]. Although the dissertation is an important part of MPH education, the studies on the quality of their dissertation are still limited [6].
Randomized controlled trial (RCT) has been advocated as the gold standard for evaluating treatment modalities. However, many studies are di cult to be veri ed by trials due to various ethical problems and side effects of intervention in practice [7][8][9]. Well-produced observational studies can not only provide abundant clue for investigating the causal relationship between exposure and diseases, but also be more suitable for investigating long-term and rare side effects of treatment modalities. What's more, one study showed that about 90% of the papers published in medical journals are observational studies [10].
Compared with RCT, the researchers of observational studies cannot randomly assign the study factors to the participants. They can only rely on comprehensive, objective descriptions or well-designed programs to analyze, compare and summarize the population phenomena, and further explore the causal relationships between disease and exposure factors. Hence, the reporting of observational studies should be transparent, clear and complete. The standardized reporting of observational studies not only can help the editors and reviewers of medical journals to better understand the research design, but also provide important information for readers in related elds, so that they can clearly understand the content and results of the research and improve their professional skills. Several incipient studies on reporting quality had recognized de ciencies in the medical education articles, but all of these were limited by incomplete reporting quality evaluation standards [11][12][13][14].
In 2008, the Strengthening the Reporting of Observation Studies in Epidemiology (STROBE) statement was published to "improve transparency in reporting of observational studies" [15][16][17][18]. A total of 22 items in the STROBE statement include evaluations of such as research design, data collection, analytical techniques, and potential deviations. Since STROBE statement published, several studies have evaluated the reporting quality of clinical medicine articles and found that the reporting of observational studies need to be improvement [19][20][21][22][23].
At present, no research on evaluating the reporting quality of the dissertation of Chinese master in medical has been found. Therefore, we used the STROBE statement to evaluate the reporting quality of the observational studies in MPH's dissertation in China, identify factors associated with high-quality reporting, and provide direction for writing of MPH dissertation.

Method
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline was observed throughout the design, performance, analysis, and reporting of this systematic review (supplement 1).

Search Strategy
We searched the relevant studies in the Wanfang database. The language was limited to Chinese, and the search strategy was (theme:("cohort studies" OR "cohort analyses" OR "case-control studies" OR "casecontrol analyses" OR "cross-sectional studies" OR "prevalence studies" OR "current situation studies") *profession:( Master of Public Health) * degree:( master)) * Date:2014-2019.

Study selection
Articles that met the following criteria were selected: (1) observational studies, including cross-sectional studies, case-control studies, and cohort studies; (2) original articles; (3) studies on humans, including both adults and children. The exclusion criteria were as follows: (1) review articles; (2) case reports; (3) quasi-randomized trial, randomized controlled trials and other interventional studies; (4) unpublished data and published abstracts only.
The articles retrieved were preliminarily reviewed according to titles and abstracts by two investigators independently. Any disagreement was resolved by consulting with a senior author. After the initial screening, the full text of relevant research was searched, and two investigators determined the nal literature based on inclusion and exclusion criteria.

Data extraction
The extraction of data was performed independently from included articles by two investigators. The general information extracted included publication time, type of study, number of papers published during master studies, funding support, and number of statistical methods used.

Quality assessment
According to the detailed per-item descriptions of the STROBE statement, appraisal of reporting was performed by two investigators. The STROBE statement contains 22 items: title and abstract (item 1), introduction (item 2~3), method (item 4~12), result (item 13~17), discussion (item 18~21), other information (item 22). A score of 1 was assigned to the items that all the information detailed was reported, a score of 0.5 was assigned to the item that the information detailed was partly reported, and a score of 0 was assigned to the item that the information was not at all reported. For items with subparts, the fractional points were scored depending on the number of sub-items met. Sub-item 6b was only applicable to match studies, 14c only applicable to cohort studies, and the sub-item were removed from the denominator if they were not applicable [24]. Therefore, every study had an overall STROBE score rated out of a maximum score of 22. The consistency between two investigators was evaluated with kappa, when κ ≥ 0.8, the average score of the two investigators was calculated as the nal STROBE Score, and when κ < 0.8, the differences will be resolved by discussion with the senior author until all differences were resolved.

Data analysis
The continuous data subjected to normal distribution was presented as mean and standard deviation (SD). Categorical variables were expressed as numbers and percentages. Comparisons of STROBE score between dichotomous groups were conducted using the independent Student's t-test. Comparisons of study quality between multigroups were conducted using the one-way analysis of variance (ANOVA), with the LSD-t test. The included articles were further divided into high and low reporting quality groups according to the cut-off value (the 75 percentile of the STROBE Score). Univariate logistic regression models were used to analyze the associations between high reporting quality and study type, publication time, papers published during master studies, funding support, and types of statistical methods used. Candidate variables with a P ≤ 0.05 on univariate logistic were included in multivariable logistic regression model. The odds ratios (OR) and 95% con dence intervals (CI) were calculated from logistic regression analyses. Statistical analyses were performed using SPSS version 18.0. All reported probabilities (P values) were two-sided, and P ≤ 0.05 was considered signi cant.

Search Results
After the search of database, we con rmed 425 articles without duplication. After the screening of titles and abstracts, 201 articles were excluded. A total of 224 full articles were further reviewed, with 59 articles excluded, because 32 articles were review studies, and 27 were intervention studies. Finally, 165 relevant articles meeting the inclusion criteria were included. Of the 165 articles, 61 articles were crosssectional studies, 66 articles were case-control studies, and 38 articles were cohort studies ( Figure.1).

The compliance of included studies with STROBE statement
Adherence with 22 items of the included studies was shown in Table 1 (Table 1). The median of compliance with STROBE statement was 67.82% (95% CI: 30.61%-95.37%; range: 0-100%). The overall reporting quality was relatively good. Five items/sub-items (14.29%) were 100% reported: background/rationale, objectives, study design, report numbers of individuals at each stage of study, and key result; 15 items/sub-items (42.86%) were reported by 75% or more of the studies. Speci cally, reporting on title and abstract section and introduction section was satisfactory: reporting rates for all four items exceeded 95%. Reporting on methods and results were often omitted: variables(n = 67, 40.61%), examine subgroups and interactions (n = 21, 12.73%), missing data (n = 21, 12.73%), sensitivity analyses (n = 5, 3.03%), reasons for non-participation (n = 14, 8.48%), ow diagram (n = 25, 15.15%), and absolute risk (n = 0, 0%). The reporting rates of each item were similar in the three study types, but the reporting of cohort studies were slightly better ( Figure.).

STROBE Score
A consistent analysis of the STROBE scores from the two investigators showed good consistency (κ = 0.861, P<0.001). Hence, the mean of score of two valuators was used as the nal STROBE score in this study. The mean of STROBE score was 14.3 (range: 10.0-18.9) with a standard deviation of 1.91. We found that the STROBE score of the cohort studies was statistically signi cant higher than the other two studies. Dissertations with funding support were more likely to get high STROBE score. STROBE score of the dissertations with more published papers during postgraduate period and more statistical methods was higher than the others. The mean of STROBE score of the dissertations published in 2018 was higher than those published in other years. The characteristics of 165 included articles were shown in Table.2 (Table.2).

Univariate and Multivariate Logistic Regression Analyses
According to the cut-off value of the STROBE Score (15.70), the included articles were divided into low (n = 122) and high reporting quality groups (n = 43). Univariate logistic regression analyses showed that the  (Table.3).

Summary of Findings
Our study evaluated 165 MPH professional dissertations. Although the overall reporting quality was relatively good, some essential aspects of methods and results were seldom reported, which made di cult for readers to assess the validity and reliability of the observational studies [16]. What's more, dissertations of superior reporting quality usually contain the following predictive factors: funding support and more published papers during postgraduate period.
Reporting on title and abstract section and introduction section was satisfactory. That may be because each master needs to go through the strictly opening and middle screening stages in the early stage of the dissertation writing. The de ciency of the reporting of MPH dissertations was mainly focused on methods and results. In particular, there was a need for dissertations to improve their reporting of variables de nition, statistical methods, and ow diagram.
In actual studies, the outcome, exposure, predictors, potential confounders, and effect modi ers of the study should be clearly de ned, but less than half of dissertations fully reported these content. The inadequate reporting of statistical method may not make full use of research results, resulting in a waste of valuable information and varying degrees of bias. However, only a few articles described any methods used to examine subgroups and interactions, explained how missing data were addressed, and described any sensitivity analysis. Only 25 dissertations (15.15%) used the ow diagram, while others did not take advantage of the simple and direct features of ow diagram. In addition, all articles summarised key results with reference to study objectives, but only about a quarter of the articles discussed the generalisability of the study results.
The result of multivariate logistic regression analysis indicated that the funding support was associated with high reporting quality. Funding projects require rigorous research designs, and need to be screened and approved. Therefore, masters are strictly required and trained to learn more knowledge, so that their thesis quality will be higher. Morever, the positive association between more published papers during postgraduate period and high reporting quality were observed. The masters published more papers during postgraduate period have stronger academic ability, are ore familiar with the writing of the articles, and know what should be reported in detail. In addition, the results of univariate logistic regression analyses showed that the number of statistical methods was associated with high reporting quality. Masters who use more statistical methods have a deeper understanding of methodology, are more pro cient in using statistical methods, and are more complete in methods reporting in dissertations.

Compared with other studies
A few articles had evaluated the reporting quality of observational studies in other medical disciplines. Several studies have found that the reporting quality of the articles that used STROBE statement to standardize was better than the others [19,20,25,26]. Jacqueline Ramke et al. has used the STROBE statement to evaluated the reporting in the blindness prevalence surveys, and they found that the mean of STROBE score of studies published in the journal requiring STROBE statement was higher than the score for the others [19]. Swords C's study indicated that STROBE statement had increased the reporting quality of observational Otology and Audiology studies [20]. Hence, we strongly recommend that masters should be familiar with STROBE statement.
On the other hand, many studies have found that reporting of observational studies have defects in methods and results [27][28][29]. Adams AD et al. discovered that poor reporting was seen in obstetrics observational studies for study size, missing data, and absolutely studies [27]. Karaçam Z have evaluated the reporting quality of observational studies in Turkish nursing journals, and found that the methods sections of the reports were mostly omitted [29]. Our research had yielded similar results.

Education implication
Our study has highlighted the important de ciencies in the reporting of observational studies in MPH's dissertations. Based on these ndings, we believe that if universities adopt the STROBE criteria to guide MPH, it will help improve the reporting quality of MPH's dissertations. In the course of master's training, it is necessary to strengthen the understanding and exible application of the statistical methods, and the graduate tutors should pay more attention to the masters who published less papers during postgraduate period.

Strengths and limitations of this study
As a systematic review of MPH's dissertations, our study has some advantages. First, this study is a comprehensive assessment and used logistic regression analyses to identify factors associated with high-quality reporting. For evaluation of dissertations, we included not only adherence with items but also STROBE score. Second, some of the evaluation items are not applicable to all articles, such as 6b, 14c, and some items are not adequately reported. To minimize biases against systematic review, we identi ed items as: full reporting, partly reporting, no reporting, not applicable, and assign different scores. Thus, different articles have a more consistent score criterion. Third, since the STROBE statement was published in 2007, no studies has used this guidelines to evaluate the reporting quality of master dissertation. Therefore, our work is innovative and will provide a reference for subsequent similar research. Fourth, the study include the independent assessment of all articles by two authors. All details of our search were transparent and clear, and our search can therefore easily be reproduced.
There are also some limitations in this study. First, scoring of items remains a subjective task, with differences between investigators. Several items were prone to discussion between the investigators. However, the analysis showed that we reached good interobserver agreement (κ = 0.861). Second, given that our research was restricted to MPH dissertations published by Chinese master in the past ve years, the results can only re ect the integrity and standardization of the reporting of Chinese MPH's master dissertation to a certain extent. Finally, since there is no literature to be found on the reporting quality of the medical master dissertation with STROBE statement, it is impossible to compare the reporting quality of the dissertation with other profession.

Conclusion
In short, the reporting quality of observational studies in MPH dissertation is suboptimal. However, there is still a need to improve the reporting of method and results sections, especially statistical methods' reporting. The STROBE statement was intended to help researchers to improve transparency in reporting of observational studies. Therefore, we think it is highly plausible that using the STROBE Statement will improve the quality of reporting. We recommend masters of observational studies to use the STROBE Statement and recommend research supervisor to use this Statement to guide MPH masters.   Figure 1 Flow diagram of the literature