Author, year, country | Study sample selection bias additional criteria for case– control studies | Bias in measurement of exposure | Bias in measurement of outcomes | Attrition bias | Analysis method reported and appropriate | Potential confounders adjusted for and listed | |
UK Studies | |||||||
Dummer & Parker 29, 2004, UK | LOW Cumbrian Births Database | LOW Modelled using GIS | LOW Objective outcome (NM) | LOW Of 3352 live births, 42 stillbirths excluded as the outcome NM | LOW Method detailed, results of LR were reported | LOW Adjusted for year of birth, social class, birth order, multiple births | |
Paranjothy et al 30 31, 2013 & 2014, UK | LOW National Community Child Health Database & All Wales Perinatal Survey) | HIGH Women’s address replaced by population-weighted centroid, travel time calculated using Google Maps API (v3) | LOW Objective outcomes (SB, NM) | LOW 11% excluded where information on parity was missing | LOW Analysis method described and multilevel LR data were reported | LOW Adjusted for maternal age, parity, urban/rural location, SES, and other characteristics | |
European Studies | |||||||
Combier et al 32, 2013, France | LOW Burgundy perinatal network database | HIGH Municipality town hall not woman’s home address | LOW Objective outcomes (SB, PM, BBA) | LOW All births identified included in the analysis | LOW Method described; hierarchical LR and multilevel LR reported | LOW Adjusted for maternal age, urbanisation level and other characteristics | |
Renesme et al 34, 2013, France | Case definition | YES linked to perinatal network database | LOW Distance & travel time estimated using GIS | LOW Objective outcomes retrieved from regional and hospital databases | LOW 5/81 (6%) BBAs missing, 3/162 (2%) controls missing | LOW Method described and univariate and appropriate; multivariate reported | LOW Adjusted for age, family status, INSEE maternal occupation, parity, and other characteristics |
Representativeness of cases | YES All cases in defined period | ||||||
Appropriate selection of controls | YES Controls chosen randomly from same databases and from births occurring at the nearest delivery date and hour to cases | ||||||
Definition of control appropriate | YES Outcome could not have occurred | ||||||
Comparability of cases and controls | NO Difference in antenatal care attendance | ||||||
Nguyen et al 33, 2016, France | Case definition | YES Using medical records | UNCLEAR No information | LOW Objective outcome (BBA) | UNCLEAR No information | HIGH No details of the analysis method and analysis was only descriptive | HIGH No adjustment for any potential confounders |
Representativeness of cases | YES All cases in defined period | ||||||
Appropriate selection of controls | YES Next birth, of equivalent GA | ||||||
Definition of control appropriate | YES Outcome could not have occurred | ||||||
Comparability of cases and controls | NO Significant differences in parity, smoking, pregnancy monitoring, profession | ||||||
Ravelli et al35–37, 2011, Netherlands (abstract & full papers) | LOW Population based study using the Netherlands Perinatal Registry | LOW GIS software used to measure travel time from women's postcodes | LOW Objective outcomes from perinatal registry | LOW Small proportion (0.3%) of women excluded due to incorrect zip code | LOW Method reported; descriptive analysis & LR results given | LOW Analysis adjusted for age, parity, ethnicity, SES | |
Ravelli et al 38, 2012, Netherlands | LOW Population based study using the Netherlands Perinatal Registry | LOW GIS software used to measure travel time from women's postcodes | LOW Objective outcomes from perinatal registry | LOW Small proportion 4% of women excluded | UNCLEAR No information | LOW Adjusted for age, parity, very urban /very rural, SES | |
Stolp et al 39, 2015, Netherlands | HIGH Study participants were selected by midwives | LOW Ambulance interval includes total time from dispatch call to arrival at hospital | UNCLEAR Method of measuring blood loss not reported | HIGH Missing data 18/72 (25%) due to incomplete documentation | HIGH Data only analysed descriptively | HIGH No adjusted analysis | |
Engjom et al 40, 2017 and Engjom et al 41, 2015, Norway (abstract & full paper) | LOW Medical Birth Registry of Norway | LOW Travel time polygon from home address using GIS | LOW Objective outcomes from birth registry | UNCLEAR No information | LOW Analysis appropriate, details of LR, multilevel modelling were reported | LOW Adjusted for maternal age, parity, education, ethnicity | |
Grzybowski et al 43, 2011, Canada | LOW Population based study using British Columbia Perinatal Health Programme | HIGH GIS used to create 1 hour travel zone for each maternity service, but central postal code to the nearest maternity care used | LOW Objective outcomes from Perinatal Health Programme | LOW 0.3% excluded due to incorrect zip code | LOW Analysis appropriate, descriptive analysis & hierarchical LR reported | LOW Adjusted for maternal age, parity, SES, ethnicity | |
Grzybowski et al 44, 2013, Canada | LOW Data from Perinatal Data Registry | HIGH Community central postal code used not women’s home address | LOW Objective outcomes from Perinatal Data Registry | HIGH Number of women excluded due to incorrect postal address not reported | LOW Analysis appropriate descriptive analysis & LR | LOW Adjusted for maternal age, parity, lone parent status, ethnicity, SES | |
Grzybowski et al 45, 2015, Canada | LOW Provincial perinatal registries | HIGH Community central point postal code used not women’s home address | LOW Objective outcomes from Perinatal Data Registries | UNCLEAR No information on missing data | LOW Analysis appropriate, descriptive analysis & LR reported | LOW Adjusted for maternal age, parity | |
Stoll et al 46, 2014, Canada | LOW Based on British Columbia Perinatal Database Registry | LOW Used GIS and Google maps; travel times were adjusted for travel conditions | LOW Objective outcomes (CS) | LOWNo missing data | HIGH Data were only analysed descriptively | HIGH No adjusted analysis | |
Darling et al 42, 2019, Canada | LOW Data from Perinatal Registries | LOW Driving time from women’s residence using online mapping tool ArcGIS | LOW Objective outcomes from Perinatal Data Registries | LOW 3.7% excluded not being able to calculate distance to nearest hospital | LOW Method reported, results of descriptive analysis & LR reported | LOW Adjusted for maternal age, parity, gestational age, season, SES | |
Aoshima et al 47, 2011, Japan | LOW All Japan except for isolated islands outside road network (96.6% of all Medical Service Areas) | HIGH Used central point of municipality not home address but analysis based on (larger) Medical Service Areas. | LOW Objective outcomes from Medical Service Area databases | UNCLEAR No information on missing data | LOW Method appropriate, unpaired t-test, difference-in-difference analysis | HIGH No adjusted analysis |
BBA, born before arrival; CS, caesarean section; GA, gestational age; GIS, geographical information system; INSEE, institute national de la statistique et des etudes economiques; LR, logistic regression; NM, neonatal mortality; NNU, neonatal unit; PM, perinatal mortality; SES, socio economic status; SB, still birth.