Accuracy of patient perceptions of maternity facility quality and the choice of providers in Nairobi, Kenya: A cohort study

BMJ Open, Volume 9, No. 7, Year 2019

Interpretation

Objectives This study aimed to assess the accuracy of pregnant women’s perceptions of maternity facility quality and the association between perception accuracy and the quality of facility chosen for delivery. Design A cohort study. Setting Nairobi, Kenya. Participants 180 women, surveyed during pregnancy and 2 to 4 weeks after delivery. Primary outcome measures Women were surveyed during pregnancy regarding their perceptions of the quality of all facilities they were considering during delivery and then, after delivery, about their ultimate facility choice. Perceptions of quality were based on perceived ability to handle emergencies and complications. Delivery facilities were assigned a quality index score based on a direct assessment of performance of emergency ‘signal functions’, skilled provider availability, medical equipment and drug stocks. ‘Accurate perceptions’ was a binary variable equal to one if a woman’s ranking of facilities based on her quality perception equalled the index ranking. Ordinary least squares and logistic regressions were used to analyse associations between accurate perceptions and quality of the facility chosen for delivery. Results Assessed technical quality was modest, with an average index score of 0.65. 44% of women had accurate perceptions of quality ranking. Accurate perceptions were associated with a 0.069 higher delivery facility quality score (p=0.039; 95% CI: 0.004 to 0.135) and with a 14.5% point higher probability of delivering in a facility in the top quartile of the quality index (p=0.015; 95% CI: 0.029 to 0.260). Conclusions Patient misperceptions of technical quality were associated with use of lower quality facilities. Larger studies could determine whether improving patient information about relative facility quality can encourage use of higher quality care. The study did not involve patients, as it was based on a secondary analysis of data collected for a previous study.12 We used data collected for a randomised controlled trial described in detail in Cohen et al (2017).12 The study took place between 2015 and 2016 in 24 neighbourhoods within the informal settlements (‘slums’) surrounding Nairobi, Kenya. These neighbourhoods are densely populated, with limited access to social services.24 Pregnant women were recruited through community recruitment events, community health worker listings and snowball sampling. For study inclusion, women had to be in their 5th to 7th gestational month, at least 18 years old, planning to stay in Nairobi until at least 2 weeks postpartum, and intending to deliver at a health facility. Women were surveyed at three time-points: baseline (5th to 7th gestational month), midline (8th gestational month) and endline (2 to 4 weeks after delivery). The baseline and midline surveys captured basic demographic information and pregnancy-related history. The endline survey captured information about the woman’s delivery, including her facility of choice. Three-quarters of the sample was randomly selected to be surveyed at baseline and midline about perceptions of delivery facility quality. Women in this subsample were asked to list all facilities being considered for delivery, regardless of how likely they thought it was that they would use the facility. Each facility name was written on a piece of paper and cut out. Women were then asked to rank these facilities on a visual analogue ladder scale from best to worst based on different dimensions of perceived quality and cost, with ties allowed. The quality dimension used in this study was women’s ranking of facilities based on her perception of their ‘ability to handle emergencies and complications’. Information about the technical quality of facilities where women delivered was also collected. The facility assessment was adapted from the Averting Maternal Death and Disability Program’s emergency obstetric and newborn care (EmONC) needs assessment toolkit.25 This toolkit assesses inputs including infrastructure, human resources, supplies and equipment for EmONC. It collects information about performance of ‘signal functions’ of basic and comprehensive emergency obstetric and newborn care, which have been shown to be correlated with delivery outcomes.25 The assessment also collects information on recent performance of the routine care signal functions proposed by Gabrysch et al and Tripathi et al. 26 27 A facility technical quality index was constructed based on the data collected from the facility assessment. The index captured a facility’s ability to handle emergencies and complications through measures of facility process, equipment, supplies and skilled provider availability (online supplementary table A1). It included 17 facility-reported signal functions measuring recent performance of emergency obstetric and newborn care practices, for example the administration of parenteral oxytocin for postpartum haemorrhage and parenteral antibiotics for newborn sepsis.26 27 The index also included a facility-reported variable for whether at least one medical officer was present onsite 24 hours a day and 7 days a week, as well as a binary variable for whether the facility was observed to have stocks of certain essential equipment necessary for common and rare complications as defined by Ngabo et al.28 The facility quality index was calculated as the fraction of these 19 variables met at each facility. bmjopen-2019-029486supp001.pdf We created a measure of accurate perceptions of facility technical quality using women’s rankings during their 8th gestational month. This variable is equal to 1 if the woman’s ranking of facilities based on her perception of their ability to handle emergencies matched the actual ranking based on the quality index. For women who had more than one facility with the same relative ranking (17% of the study sample), perceptions were considered accurate if the facilities had an identical quality index score. Five hundred and fifty-three women were surveyed at baseline. Among these, 459 women and 454 women were reached for midline and endline, respectively. Attrition was primarily due to temporary relocation around the time of delivery to be with family members, miscarriages or newborn mortality.12 The study sample was constructed from the women surveyed at midline who were also randomly selected to be asked about facility quality perceptions (n=334). We restricted the sample to women considering more than one facility, in order for us to assess their quality rankings (n=280), and to women whose consideration set included at least two assessed facilities (n=221). Finally, we only included women whose delivery facility was assessed. The final analysis sample included 180 women. The characteristics of this analysis sample are similar to the original baseline sample ((online supplementary table A2). Seventy-nine health facilities were targeted for assessment in the original study and all but 15 of these were reached for assessment (online supplementary figure A1). Incompletion was primarily due to facility administrative delays and permanent facility closure. Of these 64 health facilities assessed in the original study, 3 reported no deliveries in the past 3 months and 22 were not used by women in our analysis sample. The facility analysis sample thus included 42 facilities, of which 16 were public, 18 were private and eight were non-governmental organisations (NGO)/mission facilities. Twenty of the facilities were hospitals and 22 were health centres. The primary outcome was the quality of facility used for delivery with respect to the facility’s ability to manage emergencies and complications, as measured by the quality index. We used ordinary least squares regressions to analyse the relationship between accurate perceptions and the quality index for the delivery facility used. Adjusted regressions included indicator variables for treatment arm in the original randomised controlled trial, neighbourhood location, gestational month at baseline and the number of facilities in a woman’s consideration set. They also included: (i) a categorical variable indicating if she previously had a C-section or whether the child was a first birth, (ii) a binary variable indicating whether she reported receiving information from a health worker that her current pregnancy was high-risk, (iii) a binary variable indicating whether she reported that it would be difficult to collect 1000 Ksh, (roughly 10 US$) if needed for a health emergency, (iv) a binary variable for whether she obtained a education or higher, (v) a binary variable indicating marital status, (vi) a binary variable indicating health insurance status and (vii) a binary variable indicating whether she had four or more prenatal care visits. In ordinary least squares and logistic regression specifications, we also estimated the association between accurate perceptions and delivery in a facility in each quartile of the quality index range and in facilities at different levels of the health system (primary health centre, hospital and tertiary hospital). Robust standard errors were used in the ordinary least squares specifications. We tested the robustness of the results to a different definition of facility quality and to a more lenient definition of accurate perceptions. For facility quality, we used the number (count) of basic emergency obstetric and newborn care (BEmONC) signal functions (out of 7) performed.29 For the more lenient definition of accurate perceptions, we consider women who incorrectly ranked two facilities with equal index values as not being tied to have had ‘accurate’ perceptions.

AI Digest

Study Justification:

– The study aimed to assess the accuracy of pregnant women’s perceptions of maternity facility quality and its association with the quality of the facility chosen for delivery.
– The study is important because patient misperceptions of technical quality were found to be associated with the use of lower quality facilities.
– The findings suggest that improving patient information about relative facility quality could encourage the use of higher quality care.

Study Highlights:

– The study was conducted in Nairobi, Kenya, in 24 neighborhoods within informal settlements.
– A total of 180 women were surveyed during pregnancy and 2 to 4 weeks after delivery.
– Perceptions of facility quality were based on the perceived ability to handle emergencies and complications.
– The study found that 44% of women had accurate perceptions of quality ranking.
– Accurate perceptions were associated with higher delivery facility quality scores and a higher probability of delivering in a facility in the top quartile of the quality index.

Study Recommendations:

– Larger studies should be conducted to determine whether improving patient information about relative facility quality can encourage the use of higher quality care.
– Efforts should be made to provide pregnant women with accurate information about the quality of maternity facilities.
– Policy makers should consider implementing interventions to improve patient perceptions of facility quality and promote the use of higher quality care.

Key Role Players:

– Researchers and data analysts to conduct larger studies and analyze the data.
– Community health workers to provide accurate information to pregnant women.
– Health facility administrators to implement interventions to improve facility quality and patient perceptions.

Cost Items for Planning Recommendations:

– Research funding for larger studies and data analysis.
– Training and capacity building for community health workers.
– Implementation costs for interventions to improve facility quality and patient perceptions.
– Monitoring and evaluation costs to assess the effectiveness of interventions.

Strength of Evidence

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is moderately strong, with a rating of 7. The study design is a cohort study, which provides a higher level of evidence compared to cross-sectional or case-control studies. The study sample size is relatively small (180 women), which may limit the generalizability of the findings. However, the study used a rigorous methodology, including direct assessment of facility performance and women’s perceptions of quality. The results show a significant association between accurate perceptions of facility quality and the choice of higher quality facilities for delivery. To improve the strength of the evidence, larger studies could be conducted to confirm the findings and increase the generalizability. Additionally, including a control group or randomization could help establish a causal relationship between accurate perceptions and facility choice.

Abstract

The study did not involve patients, as it was based on a secondary analysis of data collected for a previous study.12 We used data collected for a randomised controlled trial described in detail in Cohen et al (2017).12 The study took place between 2015 and 2016 in 24 neighbourhoods within the informal settlements (‘slums’) surrounding Nairobi, Kenya. These neighbourhoods are densely populated, with limited access to social services.24 Pregnant women were recruited through community recruitment events, community health worker listings and snowball sampling. For study inclusion, women had to be in their 5th to 7th gestational month, at least 18 years old, planning to stay in Nairobi until at least 2 weeks postpartum, and intending to deliver at a health facility. Women were surveyed at three time-points: baseline (5th to 7th gestational month), midline (8th gestational month) and endline (2 to 4 weeks after delivery). The baseline and midline surveys captured basic demographic information and pregnancy-related history. The endline survey captured information about the woman’s delivery, including her facility of choice. Three-quarters of the sample was randomly selected to be surveyed at baseline and midline about perceptions of delivery facility quality. Women in this subsample were asked to list all facilities being considered for delivery, regardless of how likely they thought it was that they would use the facility. Each facility name was written on a piece of paper and cut out. Women were then asked to rank these facilities on a visual analogue ladder scale from best to worst based on different dimensions of perceived quality and cost, with ties allowed. The quality dimension used in this study was women’s ranking of facilities based on her perception of their ‘ability to handle emergencies and complications’. Information about the technical quality of facilities where women delivered was also collected. The facility assessment was adapted from the Averting Maternal Death and Disability Program’s emergency obstetric and newborn care (EmONC) needs assessment toolkit.25 This toolkit assesses inputs including infrastructure, human resources, supplies and equipment for EmONC. It collects information about performance of ‘signal functions’ of basic and comprehensive emergency obstetric and newborn care, which have been shown to be correlated with delivery outcomes.25 The assessment also collects information on recent performance of the routine care signal functions proposed by Gabrysch et al and Tripathi et al. 26 27 A facility technical quality index was constructed based on the data collected from the facility assessment. The index captured a facility’s ability to handle emergencies and complications through measures of facility process, equipment, supplies and skilled provider availability (online supplementary table A1). It included 17 facility-reported signal functions measuring recent performance of emergency obstetric and newborn care practices, for example the administration of parenteral oxytocin for postpartum haemorrhage and parenteral antibiotics for newborn sepsis.26 27 The index also included a facility-reported variable for whether at least one medical officer was present onsite 24 hours a day and 7 days a week, as well as a binary variable for whether the facility was observed to have stocks of certain essential equipment necessary for common and rare complications as defined by Ngabo et al.28 The facility quality index was calculated as the fraction of these 19 variables met at each facility. bmjopen-2019-029486supp001.pdf We created a measure of accurate perceptions of facility technical quality using women’s rankings during their 8th gestational month. This variable is equal to 1 if the woman’s ranking of facilities based on her perception of their ability to handle emergencies matched the actual ranking based on the quality index. For women who had more than one facility with the same relative ranking (17% of the study sample), perceptions were considered accurate if the facilities had an identical quality index score. Five hundred and fifty-three women were surveyed at baseline. Among these, 459 women and 454 women were reached for midline and endline, respectively. Attrition was primarily due to temporary relocation around the time of delivery to be with family members, miscarriages or newborn mortality.12 The study sample was constructed from the women surveyed at midline who were also randomly selected to be asked about facility quality perceptions (n=334). We restricted the sample to women considering more than one facility, in order for us to assess their quality rankings (n=280), and to women whose consideration set included at least two assessed facilities (n=221). Finally, we only included women whose delivery facility was assessed. The final analysis sample included 180 women. The characteristics of this analysis sample are similar to the original baseline sample ((online supplementary table A2). Seventy-nine health facilities were targeted for assessment in the original study and all but 15 of these were reached for assessment (online supplementary figure A1). Incompletion was primarily due to facility administrative delays and permanent facility closure. Of these 64 health facilities assessed in the original study, 3 reported no deliveries in the past 3 months and 22 were not used by women in our analysis sample. The facility analysis sample thus included 42 facilities, of which 16 were public, 18 were private and eight were non-governmental organisations (NGO)/mission facilities. Twenty of the facilities were hospitals and 22 were health centres. The primary outcome was the quality of facility used for delivery with respect to the facility’s ability to manage emergencies and complications, as measured by the quality index. We used ordinary least squares regressions to analyse the relationship between accurate perceptions and the quality index for the delivery facility used. Adjusted regressions included indicator variables for treatment arm in the original randomised controlled trial, neighbourhood location, gestational month at baseline and the number of facilities in a woman’s consideration set. They also included: (i) a categorical variable indicating if she previously had a C-section or whether the child was a first birth, (ii) a binary variable indicating whether she reported receiving information from a health worker that her current pregnancy was high-risk, (iii) a binary variable indicating whether she reported that it would be difficult to collect 1000 Ksh, (roughly 10 US$) if needed for a health emergency, (iv) a binary variable for whether she obtained a education or higher, (v) a binary variable indicating marital status, (vi) a binary variable indicating health insurance status and (vii) a binary variable indicating whether she had four or more prenatal care visits. In ordinary least squares and logistic regression specifications, we also estimated the association between accurate perceptions and delivery in a facility in each quartile of the quality index range and in facilities at different levels of the health system (primary health centre, hospital and tertiary hospital). Robust standard errors were used in the ordinary least squares specifications. We tested the robustness of the results to a different definition of facility quality and to a more lenient definition of accurate perceptions. For facility quality, we used the number (count) of basic emergency obstetric and newborn care (BEmONC) signal functions (out of 7) performed.29 For the more lenient definition of accurate perceptions, we consider women who incorrectly ranked two facilities with equal index values as not being tied to have had ‘accurate’ perceptions.

Materials

Innovations

The study titled “Accuracy of patient perceptions of maternity facility quality and the choice of providers in Nairobi, Kenya: A cohort study” aimed to assess the accuracy of pregnant women’s perceptions of maternity facility quality and its association with the quality of the facility chosen for delivery. The study was conducted in Nairobi, Kenya, and involved 180 women who were surveyed during pregnancy and 2 to 4 weeks after delivery.

The primary outcome measures of the study included women’s perceptions of the quality of all facilities they were considering for delivery and their ultimate facility choice. Perceptions of quality were based on the perceived ability of the facilities to handle emergencies and complications. The delivery facilities were assessed using a quality index score based on the performance of emergency “signal functions,” availability of skilled providers, medical equipment, and drug stocks.

The study found that 44% of women had accurate perceptions of the quality ranking of the facilities. Accurate perceptions were associated with higher delivery facility quality scores and a higher probability of delivering in a facility in the top quartile of the quality index. This suggests that patient misperceptions of technical quality were associated with the use of lower quality facilities.

The study highlights the importance of improving patient information about the relative quality of facilities to encourage the use of higher quality care. The findings suggest that providing accurate and accessible information to pregnant women about the quality of maternity facilities can help them make informed decisions and choose higher quality facilities for delivery.

The study was published in BMJ Open in 2019.

AI Innovations Description

The study was published in BMJ Open in 2019.

AI Innovations Methodology

To simulate the impact of the main recommendations of this study on improving access to maternal health, you could consider the following methodology:

1. Identify the target population: Determine the specific group of pregnant women in Nairobi, Kenya, who would benefit from improved access to maternal health. This could include women residing in informal settlements or areas with limited access to social services.

2. Develop an intervention: Design an intervention that focuses on providing accurate and accessible information to pregnant women about the quality of maternity facilities in their area. This could involve creating educational materials, such as brochures or pamphlets, that clearly explain the different dimensions of facility quality and how to assess them.

3. Implement the intervention: Roll out the intervention by distributing the educational materials to the target population. This could be done through community health worker visits, community events, or other channels that reach pregnant women in the study area.

4. Monitor and evaluate: Track the impact of the intervention by collecting data on women’s perceptions of facility quality before and after receiving the educational materials. This could be done through surveys conducted during pregnancy and after delivery, similar to the methodology used in the original study.

5. Analyze the data: Use statistical analysis techniques, such as logistic regression or ordinary least squares regression, to assess the association between accurate perceptions of facility quality and the choice of higher quality facilities for delivery. Compare the findings to the results of the original study to determine if the intervention has had a positive impact on improving access to maternal health.

6. Adjust and refine the intervention: Based on the findings, make any necessary adjustments or refinements to the intervention to further improve its effectiveness. This could include modifying the educational materials, targeting specific subgroups within the population, or exploring additional strategies to enhance the dissemination of information.

7. Scale up the intervention: If the intervention proves to be successful in improving access to maternal health, consider scaling it up to reach a larger population. This could involve collaborating with local healthcare providers, community organizations, or government agencies to ensure widespread implementation and sustainability.

By following this methodology, you can simulate the impact of the main recommendations of the study and assess the effectiveness of providing accurate and accessible information to pregnant women in improving their choice of higher quality maternity facilities for delivery.

Author

Dima Health

Statistics:

Citations: 6

Identifiers:

DOI: 10.1136/bmjopen-2019-029486

Research Areas:

Community Interventions, Disability, Health System and Policy, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health, Technology and Innovations

Study Countries:

Kenya

Study Design:

Cohort Study, Cross Sectional Study, randomised-control-trial

Study Approach:

Quantitative

Participants Gender:

Female