Table 2 Risk of bias of the retrospective studies
From: General vs. neuraxial anaesthesia in hip fracture patients: a systematic review and meta-analysis
Author | Bias due to confounding | Bias in selection of participants | Bias in measurement of intervention | Bias due to departures from intended interventions | Bias to missing data | Bias in measurement of outcome | Bias in selection of the reported result | Overall bias |
|---|---|---|---|---|---|---|---|---|
Basques et al. 2015 [8] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Moderate risk. |
Explanation | The authors used calculated propensity scores to mitigate the selection bias. | Comparison group was retrospectively determined according to anaesthesia technique (SA vs. GA). | The ACS-NSQIP database does not capture the type or anaesthetic dosage used. | It is a retrospective study. There were no departures from intended interventions. | No important data missing. | The database was filled with data from medical records and interviews by trained reviewers. | The authors used bivariate and propensity-adjusted multivariate regression analyses. Binary outcomes were compared using logistic regression. | |
Brox et al. 2016 [23] | Low risk | High risk | High risk | Low risk. | Low risk | High risk | Low risk. | High risk. |
Explanation | The authors used Pearson’s chi-squared test and the Kruskal-Wallis test to mitigate selection bias. | Comparison group was retrospectively determined according to anaesthesia technique (SA vs. GA). | The database does not capture the type or dose of anaesthetic used. | It is a retrospective study. There were no departures from intended interventions | No important data missing. | A hip fracture registry was used to identify the patients. No information about the people collecting the data. | The authors used a multivariable conditional logistic regression model. | |
Chu et al. 2015 [25] | Low risk | High risk | High risk | Low risk. | Low risk | Unclear risk | Low risk | Moderate risk |
Explanation | The author used calculated propensity score to mitigate the selection bias. | Comparison group was retrospectively determined according to anaesthesia technique (GA vs. NA). | The database does not capture the type or dose of anaesthetic used. | It is a retrospective study. There were no departures from intended interventions | No important data missing. | A database was used without information of the people | The author used a propensity score, Student t test, Pearson chi-square test. | |
Fields et al. 2010 [9] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Moderate risk |
Explanation | The authors used calculated propensity scores to mitigate the selection bias. | Comparison group was retrospectively determined according to anaesthesia technique (SA vs. GA). | The database does not capture the type or dose of anaesthetic used. | It is a retrospective study. There were no departures from intended interventions. | No important data missing. | A surgical clinical reviewer at each hospital collects the data. | The authors used a multivariate logistic regression. | |
Helwani et al. 2015 [10] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Moderate risk |
Explanation | The authors used a propensity score to reduce the selection bias | Comparison group was retrospectively determined according to anaesthesia technique (GA vs. NA). | The database does not capture the type or dose of anaesthetic used. | It is a retrospective study. There were no departures from intended interventions. | No important data missing. | Dedicated data personnel collect, validate and submit the data after rigorous uniform training and examination. | Demographic and clinical characteristics were compared between the two groups by using Pearson chi-square test for all categorical variables. | |
Karademir et al. 2014 [11] | High risk | High risk | High risk | Low risk | High risk | High risk | Low risk | High risk |
Explanation | Retrospective study with high risk of confounders | Comparison group was retrospectively determined according to anaesthesia technique and the surgery technique (GA vs. NA). | The database does not capture the type or dose of anaesthetic used. | No departure from intervention. | No exact data about the mortality rate in the NA and the GA group. | The data were recruited from hospital data base and patient files. | The authors used a survival analysis by Kaplan-Meier method and a cox regression model. | |
Karaman et al. 2015 [12] | High risk | High risk | High risk | Low risk | Low risk | High risk | Low risk | High risk |
Explanation | Retrospective study with high risk of confounders. | Comparison group was retrospectively determined according to anaesthesia technique (GA vs. NA anaesthesia). | The dose and type of anaesthetic used is not described. | It is a retrospective study. There were no departures from intended interventions. | No missing data. | Patient screening was performed retrospectively from hospital electronic medical record system. | The authors used the Student t-test and the Yates Continuity Correction test to compare the results between the two groups. | |
Kim et al. 2013 [13] | High risk | High risk | High risk | Low risk | Low risk | High risk | Low risk | High risk |
Explanation | Retrospective study with high risk of confounders. | The three groups (GA vs. SA vs. EA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. Especially the dose could differ between the individuals. | No departure from intervention. | No missing data. | No information about the way the results were collected. | The authors used a chi-square test, Fisher’s exact test and binary logistic regression analysis to review the results. | |
Le-Wendling et al. 2012 [14] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Moderate risk |
Explanation | The authors used calculated propensity scores to mitigate the selection bias | The two groups (GA vs. RA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. Especially the dose could differ between the individuals. | No departure from intervention. | No missing data. | The hospital service support analyst collected the data. | The authors used a multiple logistic regression model and a linear regression model to compare the result. | |
Neuman et al. 2012 [27] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Low risk |
Explanation | The author used calculated propensity scores to mitigate selection bias. | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. Especially the dose could differ between the patients. | No departure from intervention. | No missing data. | The results were collected in the New York State Inpatient Database which was overseen by the U.S. Agency for Healthcare. | The author used the Wilcoxon rank sum test and the chi-square test to compare the results. | |
Neumann et al. 2014 [15] | Low risk | High risk | High risk | Low risk | Low risk | High risk | Low risk | High risk. |
Explanation | The authors used near-far matching, standardized differences, across-hospitalmatch and a within-hospital match to reduce the selection bias. | The two groups (GA vs. EA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. | No departure from intervention. | No missing data. | No information about the way the results were collected | The authors used an instrumental variable method, the McNemar test and the x2 statistic to compare the results. | |
Patorno et al. 2014 [16] | High risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | High risk. |
Explanation | Retrospective study with high risk of confounders. | The three groups (GA vs. NA, GA + NA) were retrospectively determined according to the anaesthesia technique. | The data does not capture the dosage and type of anaesthetic used. The does could differ between the patients. | No departure from intervention. | No missing data. | The authors used the Premier research database. The data were collected from member hospitals through Premier’s informatics products. | The authors used a multi-variable logistic regression to compare the results. | |
Rashid et al. 2013 [17] | High risk | High risk | High risk | Low risk | Low risk | High risk | High risk | High risk |
Explanation | Retrospective study with high risk of confounders. | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. The does could differ between the patients. | No departure from intervention. | No missing data. | Unclear how the data were collected and how the clinical measurement was done. | The authors use SPSS version 19 for statistical analyses. However no information is given on the type of analysis | |
Seitz et al. 2014 [19] | Low risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | Moderate risk. |
Explanation | The authors used calculated propensity scores to mitigate the selection bias. | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. The does could differ between the patients. | No departure from intervention. | No missing data. | The used data sets were linked using unique, encoded identifiers and analysed at the Institute for Clinical Evaluative Sciences (ICES). | The authors used the Wilcoxon rank-sum test and chi-square test to compare the results. | |
Shih et al. 2010 [18] | High risk | High risk | Low risk | Low risk | Low risk | High risk | Low risk | High risk. |
Explanation | Retrospective study with high risk of confounders. | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The measurement of intervention is well-defined. | No departure from intervention. | No missing data. | Unclear how the data were collected and how the clinical measurement was done. | The authors used Student t-test, X2 or Fisher exact test and logistic regression to compare the results. | |
Sevtap et al. 2013 [20] | High risk | High risk | Low risk | Low risk | Low risk | High risk | Low risk | High risk. |
Explanation | Retrospective study with high risk of confounders. | The three groups (GA vs. SA vs. EA) were retrospectively determined according to the anaesthesia technique. | The measurement of intervention is well-defined. | No departure from intervention. | No missing data. | This is a retrospective study. And all the data were obtained from the medical data. | The authors used the one-way analysis of variance test for normally distributed data and the Kruskal Wallis test for abnormally distributed data. The categorical variables were compared using the chi-square tests. | |
Tung et al. 2016 [24] | Low risk. | High risk. | High risk. | Low risk. | Low risk. | High risk. | Low risk. | High risk. |
Explanation | The authors used calculated propensity scores to mitigate the selection bias. | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. The dose could be different between the patients. | No departure from intervention. | No missing data. | Data were collected in the National Health Insurance research database. No information about the persons collecting the information. | The authors used a generalized estimation equation logistic regression model and propensity score. | |
White et al. 2014 [21] | High risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | High risk. |
Explanation | Retrospective study with high risk of confounders. | The two groups (GA vs. SA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. | No departure from intervention. | No missing data. | Data were collected by specially trained personnel employed by each eligible hospital. | The authors used a two-tailed chi-squared test without Yate’s correction and multivariable regression analysis. | |
White et al. 2016 [26] | High risk. | High risk | High risk | Low risk | High risk | Low risk | Low risk | High risk |
Explanation | Retrospective study with high risk of confounders. | The two groups (GA vs. SA) were retrospectively determined according to anaesthesia technique. | The study described the different volumes used for intrathecal injections. However the dose and type for the general anaesthesia or the peripheral nerve block were not described. | No departure from intervention. | 16.904 patient records. However only 11.085 could be analysed. | Data were collected by specially trained personnel employed by each eligible hospital. | The authors used Fisher’s exact test, chi-squared, Wilcoxon and Haenzel tests. | |
Whiting et al. 2015 [22] | High risk | High risk | High risk | Low risk | Low risk | Low risk | Low risk | High risk. |
Explanation | Retrospective study with high risk of confounders | The two groups (GA vs. NA) were retrospectively determined according to anaesthesia technique. | The study does not describe the dose and type of anaesthetic used. The dose could be different between the patients. | No departure from intervention. | No missing data. | Data were collected at each hospital directly from patients medical records through risk-assessment nurses trained as Surgical Clinical Reviewers (SCR) | The authors used chi-square, Fischer’s exact test and multivariate models. |