Skip to main content

Table 4 Potential bias and efforts to address potential sources of bias (based on the Critical Appraisal Skills Programme – CASP)

From: The global maternal sepsis study and awareness campaign (GLOSS): study protocol

Potential Bias

Efforts to address potential sources of bias

Selection bias

Because this study is facility-based, estimates of the frequency of antenatal or postpartum infections exclude women who become ill in the community and do not seek care. Therefore, the study will underestimate the prevalence in the population by missing women who do not reach facility, including women with uncomplicated infections who seek treatment in the community and also severely ill women who died in the community.

Incidence estimates will note this limitation. Because the primary purpose of the study is to assess the proposed sepsis definitions which will be used largely for sepsis identification in facilities, a facility-based cohort is considered acceptable.

Broad identification criteria will be used to identify eligible women admitted to the facility in order to reduce the risk of non-inclusion of less severe cases of infection requiring hospitalisation.

In an effort to cover severe cases not presenting to the facility, data from civil registries (if available) will be obtained on the number and causes of maternal deaths among women living in the study geographical area during the study identification week.

Exposure measurement bias

Comparable measurements of health practices and outcomes given variations in the criteria used for diagnosis or management of women and neonates.

Indicators of outcomes and practices will be standardised wherever possible and defined in the Manual of Operations, and pretested. The prospective identification of eligible women will allow standardisation of the population and definitions before data collection and maximise the comparability of findings across different units and countries.

Outcome measurement bias

Incomplete follow up or a follow up that is too short. Women included in the study and their neonates will not be followed-up in the community after discharge from hospital.

Data will not be collected on long-term outcomes. Incomplete evaluation of potential benefits and harms to the fetus or neonate related to maternal infections and management strategies if fetal (complications, malformations, death) or neonatal outcomes (birth) occurs after the follow-up period

This potential bias is controlled by the study design and the hypothesis of the study is that the study period represents a typical week for all regions and facilities within the geographical region, regarding the number and characteristics of births, subjects returning to a health facility after initial discharge from hospital and the cases of maternal and neonatal sepsis. This strategy will allow to increase participation of countries and facilities by minimizing the burden of data collection while gathering key information for the development and implementation of better strategies for the prevention, identification and management of maternal and neonatal sepsis globally.

Given resources and time constraints follow-up of women and their babies after hospital discharge is not possible. Other study designs are needed to collect information on maternal and childhood outcomes after hospital discharge.

Incomplete identification of confounding factors or effect modifiers

Efforts will be made to identify and collect minimal information on key confounding or effect modifiers factors as listed above, including medical factors but also social characteristics that might affect the probability of admission to specific types of hospitals or management

The study is underpowered or unable to generate precise estimates

Efforts are made to include a large number of countries, and to select geographical areas with adequate institutional birth rates. The implementation of the modified protocol in voluntary individual facilities will provide an opportunity to increase the number of included women.

Important proportion of births occurring outside the participating facilities

This is a facility-based study, and the assumption is made that severe cases will reach the health system. The estimated number of childbirths that took place in the geographical area during the study period will be obtained to better assess the impact of the study design.

Difficulties to follow-up of women/babies

Inclusion of all facilities in the same geographical area should facilitate follow-up of women and their babies transferred

Seasonal variations in conditions leading to sepsis

Daily variations in maternity unit activities, workload

The inclusion of a large number of countries will allow expanding the geographical variability of the study. This will also limit the effect of geographic or seasonal clusters of infectious morbidities that might affect the outcomes of the study. Most of the expected cases of maternal sepsis occur in the postpartum period and are related to maternal genital tract infections, or other morbidities (e.g. urinary tract infection) not subject to seasonality. However, seasonal variability is well known for some infections that could lead to maternal sepsis (e.g. influenza, H1N1, malaria, chikungunya, chicken pox). The impact of these variations in our study is difficult to predict, but we think that we can reduce the effect of seasonality by including countries in different regions of the globe.

Inclusion of cases over a week minimizes variability of events across the days of a week (e.g. planned inductions or caesarean sections certain days of the week, less staff during the weekend, women coming to the facility at specific times).

Incomplete medical records

As part of the data quality process a random sample of forms will be will be cross-checked against medical records.

Data collection will allow investigators to inquire with the clinicians about any missing or unclear information in the medical records.

Staff in participating facilities will be informed before initiation of the study about the importance of complete and accurate history taking and medical record keeping.

Potential behaviour modification of health care providers or women due to participation in the awareness campaign preceding and during the survey

Behaviour modification due to being aware of the study and effect of the awareness campaign are possible. This potential bias is inherent to any prospective study.

The awareness campaign may have a positive impact on data collection by increasing the number of cases that could be identified during the identification week. It is however unlikely that health care providers behaviours and facility resources will be modified to better identify and treat sepsis in the short period of time between the campaign and data collection.

  1. CRITICAL APPRAISAL SKILLS PROGRAMME (CASP): Making Sense of Evidence 12 Questions to Help You Make Sense of a Cohort Study. Accesible at http://media.wix.com/ugd/dded87_7e983a320087439e94533f4697aa109c.pdf [32] (Accessed 2 Nov 2017)