Determinants of behaviors influencing implementation of maternal and perinatal death surveillance and response in low‐ and middle‐income countries: A systematic review of qualitative studies

Maternal and Perinatal Death Surveillance and Review (MPDSR) can reduce mortality but its implementation is often suboptimal, especially in low‐ and middle‐income countries (LMICs).


| INTRODUC TI ON
Maternal and Perinatal Death Surveillance and Response (MPDSR) is regarded as an important intervention to reduce maternal and perinatal mortality and is thought to have contributed to achieving this aim in several countries including India and Sri Lanka. 1 Therefore, widespread implementation of MPDSR is recommended by the World Health Organization (WHO). 2,3Although 85% of low-and middle-income countries (LMICs) have a national policy to review all maternal deaths, fewer than half are implementing MPDSR as per WHO guidelines. 4 implemented properly, maternal and perinatal death reviews can reduce maternal mortality by up to 35%, 5 and perinatal mortality by 30%. 6However, MPDSR often fails to achieve these improvements.In a survey of health facilities in four African countries, fewer than half could provide evidence of any changes resulting from MPDSR. 7MPDSR sometimes even led to unintended harmful outcomes such as worsening staff shortages or inappropriate referrals of severely ill patients, in order to avoid responsibility. 7Studies on barriers and enablers in several contexts have been emerging since the inception of MPDSR, but there is clearly a need to improve implementation of MPDSR to achieve its potential impact beyond outlining such factors.
Although behavioral science is crucial in this endeavor, there has been little research on behavioral determinants influencing the implementation of MPDSR, for example what motivates health staff, and how to improve leadership skills. 8,9To date, only one intervention to improve the implementation of MPDSR (including training, supervision, and provision of resources) has been rigorously evaluated in a cluster-randomized controlled trial. 5The only component currently being scaled up globally is "training of trainers".
We conducted a systematic review of qualitative studies that documented stakeholders' experiences of implementing MPDSR in LMICs.Our first paper used a realist lens to analyze the contexts and mechanisms underlying both the functional action cycle of successful MPDSR and the dysfunctional vicious cycle of ineffective MPDSR. 10This second article aims to understand and map the key behavioral determinants of MPDSR implementation, and from these, develop program theory for an intervention to improve its implementation in LMICs.

| MATERIAL S AND ME THODS
We conducted a systematic review of qualitative studies regarding the implementation of MPDSR in LMICs, which we report following ENTREQ guidance. 11The protocol was registered on PROSPERO (https://www.crd.york.ac.uk/prosp ero/displ ay_record.php?Recor dID=271527).

| Search strategy
We searched seven databases from inception to June 2022: CINAHL, MEDLINE, EMBASE, ProQuest Dissertations and Theses, Global Index Medicus, Web of Science, and Google Scholar using key terms for maternal or perinatal death reviews and qualitative studies (Table S1).

| Eligibility criteria
We included qualitative studies regarding implementation of MPDSR or any form of maternal/perinatal death review in LMICs.
We excluded studies in high-income countries, those solely about "near-miss" reviews, and studies with insufficient or poorly reported qualitative data.

| Study selection
Two reviewers independently screened titles, abstracts, and selected full texts against the inclusion criteria.Disagreements were resolved by discussion with a third reviewer.

| Data extraction and analysis
Studies were imported into Nvivo. 12Data on findings (themes, quotes, and other author observations) were identified by repeated reading of text especially in the results and discussion sections.
Two reviewers used a framework approach to identify and code behavioral determinants of implementation of MPDSR using the Theoretical Domains Framework (TDF) and the COM-B (Capability, Opportunity, Motivation) behavior change wheel. 13,14,15"Capabilities" were defined as knowledge and skills needed by individuals to implement MPDSR, while "opportunities" refer to all factors (physical and social) outside the individual needed to implement MPDSR."Motivation" includes factors that energize individuals to implement MPDSR, both automatic (habitual processes and emotional responses) and reflective (conscious, analytical decision making). 15ese key capabilities, opportunities, and motivational factors were used to determine the guiding principles for a complex intervention to improve the implementation of MPDSR.Guiding principles, a key part of the Person-based approach to developing complex interventions, highlight how the intervention will address issues crucial to engagement. 16Components of such an intervention were planned based on key examples of good practice and suggestions for improving implementation, extracted from the primary papers.
behavioral factors, implementation, intervention planning, low-and middle-income countries, maternal and perinatal death surveillance and review (MPDSR), qualitative, systematic review

| Quality assessment
We used the Critical Appraisal Skills Program (CASP) tool for qualitative studies to appraise the quality of full-text articles. 17

| Study selection
We identified a total of 5137 studies after de-duplication (Figure 1).Of these, 134 were assessed in full text, of which 76 were excluded.
The commonest reasons for exclusion were focus on other phenomena (for example, near-miss reviews) or lack of qualitative methods or qualitative data.

| Study characteristics
After screening, we included 59 studies, reported in 58 papers from 30 LMICs, 1,7,9,18-72 which included over 1891 participants, most of whom were health workers in hospitals and health facilities, although nine studies included national-level leaders of MPDSR and seven included community members (Table S2).Almost all the studies collected data using individual interviews and/or focus group discussions.Ten observed death review meetings and six also reviewed reports and other relevant documents.Most studies used thematic analysis although two used framework analysis, one used conversational analysis and 12 did not specify their analytical method.The majority (34) focused on maternal deaths, 19 included both maternal and perinatal deaths, and six covered solely perinatal or neonatal deaths.Several of the papers in the review reported improved outcomes although only one was nested in a randomized controlled trial which clearly demonstrated an improved outcome. 38

| Methodologic quality
All studies were of sufficient quality (Table S3).The qualitative methodology, research design, recruitment strategy and data collection were adequately described in almost all studies.However, the data analysis was unclear (not adequately described) in seven studies and inadequate in two, and most did not adequately consider the relationship between the researcher and the participants.

| Synthesis: Behavioral determinants of impactful MPDSR
Implementation of MPDSR is complex because it involves stakeholders at every level.Some behavioral determinants affect several different groups of stakeholders, while others may only affect one group (Figure 2, Tables 1 and 2; Tables S4-S7).

| Capability
The capabilities required increase cumulatively from community to health facility and leadership level (Figure 2).All stakeholders require a basic understanding of the purpose of MPDSR.Some leaders may misinterpret it as a tool for disciplining staff, 71 resulting in a well-justified fear of blame. 52All health workers need knowledge of clinical protocols and good record-keeping skills so that committees can access the information needed to identify cause of death and avoidable factors. 38,45,48,52Data collectors need specific skills on completing relevant forms, and interviewing/verbal autopsy where relevant. 1,39MPDSR committee members need additional knowledge on cause of death classification, 50,53,62,71 and skills in teamwork, audit, 41 communication (expressing disagreement without causing acrimony), 37,57,70 and making SMART recommendations. 47,48airpersons and leaders also need skills in leadership, 21,38,39 chairing, 37 maintaining confidentiality, 48 coaching, 42 and budgeting. 33ntors/supervisors of the leaders need additional mentorship skills. 38In several contexts, teams only had experience of reviewing maternal deaths and expressed a need for specific training on reviewing perinatal/neonatal deaths. 71,72

| Opportunity
Opportunities that enable implementation of MPDSR are summarized in Table 1.Social opportunity for an open and honest discussion of deaths and avoidable factors is of paramount importance at all levels.This requires strict maintenance of a "no-name, no-blame" policy and confidentiality. 1 This can be difficult to achieve in health facilities with low staff numbers, where health workers can easily recognize who was involved in management of a case. 61A safe learning environment can foster constructive dialogue, overcome barriers of hierarchy, and encourage all staff to identify errors and gaps in care, 9,32,48,51,52 even if anonymization is not possible.Conversely, a "blame culture" and hierarchical relationships stifle open discussion and result in blame-shifting rather than identifying avoidable factors and accepting responsibility. 23,33,37,40,46,48 the community, the social opportunity to collect information depended on respect of cultural norms and traditions. 20Patients and bereaved families need the opportunity to make complaints about care, 68 and their perspective could help the MPDSR process.
Paradoxically the absence of a complaints procedure pushes families to seek legal action as they see no other avenue. 68Community review meetings provide a unique opportunity to openly discuss issues which otherwise would not be discussed. 32rprisingly, key stakeholders are often unaware of recommendations addressed to them, 18,21,24,[39][40][41] so they cannot implement them.When key stakeholders responsible for implementation are not present at review meetings, they need to be informed about the recommendations.For leaders, the social opportunities to implement MPDSR were enhanced by integration with other public health programs at all stages of the process. 19,28ysical opportunity for implementing MPDSR depends on availability of reporting systems, medical records, and resources.
Although comprehensive reporting of deaths is the foundation for MPDSR, few LMICs have a robust vital registration system. 1  where there is no system for reporting deaths outside of government health facilities. 42,48,53,66,72Good medical record systems are essential for finding information on quality of care.Inadequate filing systems and missing records prevent further analysis of cases, 21,34,48 whereas lack of secure storage enables falsification of records when a death is being investigated. 48,52The review itself should be recorded on a form, which can facilitate the process if well-designed, 38 this form often being the main focus of review meetings. 52However, the requirement to complete it can hamper the review if forms are unavailable, 1,33 not anonymous, 40 too long, 1 or miss out information (such as social factors, quality of care, and recommendations). 48,50ailability of resources affects implementation of MPDSR at all levels.Health facilities require staff time to investigate cases and attend meetings, 7 as well as funding for training and implementing recommendations. 17,33Where staff are expected to work or meet outside normal working hours, some expect extra pay. 35,70Effective supervision requires the time of senior experts and their travel to relevant health facilities. 38Involvement of communities requires additional staff time and transport to conduct interviews and meetings and respect for traditions such as paying condolences. 1,18,20,66

| Motivation
The factors influencing motivation to implement MPDSR are summarized in Table 2.The most important is to uncouple MPDSR from fear of blame and negative consequences (such as disciplinary action and litigation), which motivate stakeholders at all levels to disengage from MPDSR.Both community members and health workers feared that they could be jailed or convicted by the police if they were found responsible for a death. 19,30,68,73Health workers also feared that they could be subject to disciplinary procedures, 56,69,71 punishments, 18,46 or litigation, 1,48,65,68 or required to pay compensation to family members. 40Leaders feared missing targets and put pressure on clinicians not to report maternal deaths. 46Some terms such as "negligence" and "audit" also elicited negative emotions. 18,48flective motivation came from stakeholders believing that there would be positive consequences such as a useful learning experience, 9,25,41 and that they were capable of making positive changes, 27 which would improve quality of care and reduce mortality. 22,48The desire for incentives was frequently mentioned, especially for members of MPDSR committees to attend meetings.
Staff often expected refreshments 9,39 or financial incentives, 56,61,70 but these were usually dependent on time-limited external donor funding.Withdrawal of incentives was a strong demotivator and resulted in meetings ceasing. 56Inclusion of MPDSR as an indicator for performance-based financing may be a more sustainable incentive but was only reported in one study. 66Members became demotivated when no positive changes were observed, 9,49,52 the same recommendations were often repeated, 21,22,41 there was no support, 48 no feedback of recommendations, 18 and no incentives. 22,52tomatic motivation to engage in MPDSR resulted from institutionalization of the process, such that it became part of the professional role and routine activities of health workers. 19,28Involving stakeholders in formulating recommendations motivated them to take ownership and responsibility for implementation. 9,48is was reinforced by providing feedback about implemented changes and supportive supervision. 23,27,28,42,52Health workers were automatically motivated to improve their quality of care when they knew that this would be audited as part of MPDSR. 41,47,48DSR commonly elicited negative emotions such as fear and guilt and "rebranding" was used to avoid this (e.g. from "audit" to "review"). 18,48The feeling of guilt sometimes led to defensiveness, 48 but sometimes motivated improvements in care. 62,70

| Guiding principles for an intervention to improve implementation of MPDSR
These principles follow logically from the behavioral determinants identified in Table 3.
Capability to implement the various components can be built through training, addressing specific needs in each stakeholder group.The training should be available on an ongoing basis for new staff, especially in contexts where there is frequent turnover. 19Ongoing mentorship and supervision are also necessary to continually improve capabilities. 2,19,27,28,32,38,42,43,48cial opportunities for meaningful and productive discussions can be increased at the local level by asking committee members to sign a charter, 35,70 committing themselves to observing the principles of MPDSR such as confidentiality and "no-name, no-blame", and ensuring a safe learning environment.In addition, its principles need to be enforced by the chair of meetings, which can be particularly challenging in small health facilities where staff can easily recognize themselves in case discussions. 61Good communication of recommendations is essential to ensure that those responsible have the opportunity to implement them.
Physical opportunities to implement MPDSR can be improved by ensuring data quality, such as integrated and user-friendly death reporting systems, 52 structured medical records, 39 secure and organized filing of medical records, 48 optimized MPDSR forms, and structured supervision forms. 38It is equally important to ensure that resources are sufficient, by embedding MPDSR into routine health services and ensuring that funds are available for necessary expenses such as stationery and transport, as well as implementing recommendations.
Fear of blame, disciplinary action, and litigation, as the critical issue affecting motivation, needs to be addressed at all levels, through structural changes such as preventing the use of MPDSR documents for litigation, 74 and separating responsibility for MPDSR from disciplinary procedures. 48The focus must be on recommending health system improvements rather than identifying TA B L E 3 Guiding principles for an intervention to improve implementation of MPDSR in the community (organized by behavioral determinants).

Capability Understanding purpose of MPDSR
To ensure all stakeholders (involved in MPDSR-as per Figure 1) understand the purpose of MPDSR and its core principles (including "no name, no blame" and identifying areas for collective action rather than identifying individuals) •

Knowledge of clinical guidelines and standards
To improve identification of areas where care can be improved

Opportunity Social opportunity for an open and honest discussion of deaths and avoidable factors
To enable all relevant stakeholders to identify issues with quality of care and to contribute to the discussion • Committing to confidentiality and "no-name, no blame", by signing an MPDSR "charter" • Providing a safe learning environment • Providing a feedback/complaints procedure for patients and bereaved families to provide direct feedback and make complaints • Holding community meetings to enable community members to discuss relevant cases

Social opportunity for implementation of recommendations
To ensure that all relevant stakeholders are aware of recommendations • Where possible, assign responsibility for implementation of each recommendation to a specific person • For broader recommendations, identify key stakeholders according to their influence and interest in the topic • Establish a communication and dissemination plan for communicating recommendations to all who have the possibility to implement them (especially if they are not present at the meetings)

Social opportunity to interview bereaved relatives
To respect cultural traditions around bereavement, burials, and mourning • Respect of local customs and traditions should be prioritized over other considerations (e.g.completing interview within a certain timeframe) • Appropriate condolence gifts should be provided where this is a cultural expectation

Death notification system
To streamline and facilitate death reporting • Integrated and simplified death notification system, so that community members and health workers can easily and quickly report all maternal and perinatal deaths.
individuals at fault.Rebranding may be needed in contexts where terms like "audit" and "negligence" have become associated with blame. 18,48Reviews can be used as a positive mechanism for prioritizing modifiable factors and exonerating staff from unfair blame. 65 parallel, motivation of health workers will be automatically increased if implementation becomes part of their professional role and becomes embedded in their working schedules.Involving relevant stakeholders in review meetings and formulating recommendations will enable them to take ownership and responsibility for implementing them. 9,25,27,28,38,39,48,49It is also crucial to have a system to follow-up and monitor the implementation of recommendations. 23,27,48,53flective motivation to engage in MPDSR should be increased by maximizing learning opportunities, building self-efficacy of members, and providing incentives.Most health workers are keen to keep learning and many value MPDSR meetings for this reason. 9,25,27,41eir self-efficacy can be built by experiencing improvements due to MPDSR and receiving positive feedback about recommendations implemented. 27,53Although financial compensation for participation in meetings has been offered by some projects, this often depended on donor funding and so was unsustainable, 52 causing demotivation and even collapse of the process when incentives were withdrawn. 72her more sustainable incentives include performance-based financing, 66 providing refreshments during meetings, 39 and celebration of achievements. 9

| DISCUSS ION
The principal behavioral determinants of MPDSR include capability to perform the tasks required by different stakeholders, the physi- We conducted a comprehensive literature search and included articles from a wide range of LMICs with over 1891 participants, which provides solid empirical evidence on which to base the analysis.Although the relative importance of determinants varies in different contexts, the main factors were remarkably consistent in all the studies.We did not conduct a formal assessment of confidence in the review findings.The search was limited to qualitative studies; quantitative studies may also provide useful complementary evidence and could be reviewed subsequently.

| Implications for policy and practice: proposed components of a complex intervention to improve implementation of MPDSR
Based on our findings, a complex intervention to improve MPDSR implementation could consist of six major components (Table 4).Simply implementing training in the context of a "blame culture", inadequate data quality, and lack of resources are unlikely to achieve the desired impacts.For this reason, the first four components lay the foundations on which subsequent training and supervision can be built.

Stakeholder engagement and implementation research
Engagement at all levels is key to ensure ownership of the intervention, and that results of MPDSR are transformed into concrete actions by relevant stakeholders.An implementation research

Reflective motivation to engage
To provide a useful learning experience • Ensure that MPDSR meetings provide valuable learning opportunities for all staff.
To build self-efficacy   75 Structural changes to reduce fear of blame National level legal protection for MPDSR, to prevent it from being used in litigation, has already been enacted in South Africa. 74If the system has already acquired a negative reputation inciting fear of blame, "re-branding" may be needed-for example in the Democratic Republic of Congo "audit" was replaced by "review". 48

Mobilizing resources for implementation
Based on the findings of the implementation research, implementers will define funding priorities and plan the budget for MPDSR activities, taking into account available resources at local, sub-national, and national levels.A process should be developed for considering recommendations from reviews as part of the prioritization of health spending.This would help to avoid the situation where recommendations are not implemented because they require huge investments. 56mmittees should also be encouraged to make recommendations that are actionable within available budgets or to identify resources for implementing them. 1,19,27,43 t would be most sustainable to identify funds from national and/or district budgets, rather than relying on external donors. 43modular "whole institution" approach to training and institutionalization A modular approach would be most efficient and would ensure that specific stakeholders are empowered with specific skills relevant to their level, develop a positive attitude to MPDSR, and feel responsible for the results (Figure 2 and Table 4).This "whole institution" approach has been successfully piloted to improve provision of family planning services. 76Lower levels are provided as in-service training, on-site, to minimize disruption to service delivery.Higher levels, involving smaller numbers from each facility, may be most efficient if conducted off-site.Supervisors would benefit from specific mentoring training to maximize their effectiveness, such as a package successfully used in Sierra Leone. 77

Community engagement
Involving community members in reporting, investigating, and reviewing deaths will maximize the potential impact of MPDSR, but will also require more resources. 18Therefore, this final component may be best added after the MPDSR process is already running effectively in health facilities, and when sufficient resources have been identified.
Developing an effective complaints procedure for patients and bereaved families should be a priority and may help to reduce litigation, if families feel that their feedback is acknowledged and acted upon. 68

| Priorities for future research
Detailed procedures for each of these components need to be cocreated with relevant stakeholder groups.A global "toolkit" of intervention components and resources, which could be adapted to different contexts in different countries, would save time and effort, rather than starting from scratch in each setting.These should be piloted on a small scale to refine and optimize each component, based on feedback from the target population. 16Where good internet is available, some components could efficiently be delivered online (such as the WHO virtual training course), 78 whereas more face-to-face training will be required in areas with poor connectivity.Process and effectiveness evaluations of the intervention will help to improve it iteratively.
It is very important to evaluate the cost-effectiveness of the intervention package and of MPDSR itself.Although many countries have no budget allocation for MPDSR, 7,16,26,48 adequate resources (staff and materials) are essential to achieve good results. 1 This has an opportunity cost as well as a financial cost, and it has been argued that resources would be better spent implementing interventions with proven effectiveness. 79However, an effective MPDSR process is itself a tool to improve implementation and uptake of "proven interventions", 27,32,48 by changing behavior and prioritizing use of scarce resources. 1rther qualitative research is needed to understand the views of health workers and communities on how to achieve the optimal balance between "no blame" and "accountability".There is an ethical imperative to safeguard vulnerable patients from deliberate harm and negligence, so absolute confidentiality can never be guaranteed.Further research is needed to find the optimal ways of separating disciplinary procedures from MPDSR, especially when the same leaders are in charge of both.Research is also needed to optimize involvement of communities and bereaved families in MPDSR, so that this process helps to address their concerns and empower them to take appropriate actions to avoid future deaths, without engendering fear of blame in either the families themselves or health workers.

| CON CLUS IONS
Implementation and impact of MPDSR could be improved by (1) engaging key stakeholders in an "implementation research" approach, (2) introducing structural changes to reduce fear of blame, (3) improving data collection tools and information systems, (4) mobilizing adequate resources, (5) building the capabilities of all stakeholders, and (6) community involvement.These strategies would address the major behavioral determinants which influence implementation of MPDSR in many LMICs.
Abbreviation: MPDSR, Maternal and Perinatal Death Surveillance and Review.a Blue = physical opportunities; pink = social opportunities.
Community meetings to explain MPDSR and address any concerns • Pre-service training: incorporation of basic training in curricula for all health workers, especially doctors and midwives • In-service training for all health workers • Refresher training: Regular updates for all health workers and to ensure that new health workers are also trained • Provide opportunities for all staff to be involved in regular MPDSR meetings Documentation and record-keeping skills To improve clinical record-keeping • Persuade and train health workers on importance of comprehensive record-keeping • Improve vigilance through auditing of records Data collection skills To improve data collection • Training on completing relevant forms • Training on verbal autopsy/interview techniques (for data collectors) Knowledge of death and cause of death classification To improve accuracy of death and cause of death classification • Training on basic classification of maternal and perinatal death types (what counts as a maternal death, stillbirth vs neonatal death) • Training on ICD-MM (maternal mortality) and ICD-PM (perinatal mortality) cal and social opportunity to conduct reviews and implement their recommendations, and the automatic and reflective motivation to engage in the process.Based on this empirical evidence, guiding principles for an intervention to improve implementation of MPDSR include building capabilities at all levels, improving opportunities for successful MPDSR (by improving data quality and availability, mobilizing resources and creating a learning environment), and motivating all stakeholders to engage in the process.Motivation requires removing fear of blame and can be increased automatically by embedding MPDSR into institutions and professional roles, involving all important stakeholders, and establishing systems for monitoring implementation.Motivation can be enhanced by providing valued learning opportunities, building self-efficacy of committee members, and providing context-specific incentives.

•
Provision and training on relevant evidence-based guidelines and standards for both maternal and perinatal care • Use of structured approach to discussion; training on use of fishbone diagrams.• Ensure that meeting is confidential and anonymous • Skillful chairing of meetings to ensure blame-free process and to facilitate all members to be appropriately self-critical • Input from external reviewers • Regular supervision by experienced mentors • Training on coaching/training/mentorship skills, tailored to each level of leadership as appropriate • Use of cheap and widely used communication channels (e.g.
Assessment of health facilities' readiness to implement MPDSR Interviews and focus group discussions with key stakeholders Removing fear of blame Legal protection Enactment of legal instruments to prevent use of MPDSR data in litigation