A Mixed-Methods Assessment of the Maternal Death Surveillance Response in Tanzania as an Information Source Beyond the Numbers: An Opportunity for SDG 2030 Tracking

Abstract

This study aimed to assess the implementation of the MDSR system. The mixed cross-sectional study was conducted in 12 regions. The quantitative analysis used ordinal intervals: Excellent, Good, Satisfactory and Poor. Qualitative analysis developed themes to inform MDR system implementation. The findings showed that leaders were following up at the health facility, the community, and individual clients. The system implementation at the regional level was Satisfactory (75–84%), and at the district and health facility level was Poor (below 75%). Generally, responses to the action plan at all levels studied were rated Poor. The timeliness of notification to MD was found to be Satisfactory (75–84%) in three regions and Poor in Dar es Salaam and Mwanza. The daily notification of MDs was 57.3% of the 349 weekly reports. The completeness variables improved significantly to Excellent scores (above 95%) after a week, but not for notifications. Therefore, the system produces data that goes beyond numbers, thereby making it not inferior. The system generates MD data through timely notifications and weekly reporting, with the latter yielding twice as much.

1. Introduction

Maternal mortality remains a critical global health challenge and is a key indicator of health system performance. Sustainable Development Goal 3 (SDG 3), which calls for a substantial reduction in maternal mortality by 2030, aligning with the RMNCAH Global Strategy for Women and Children’s Health (2016–2030) [1,2]. Achieving this requires reliable, timely, and actionable data to guide interventions and monitor progress.

In Tanzania, maternal mortality has traditionally been estimated using the sisterhood method. Although widely applied, this approach is limited by wide confidence intervals due to small sample sizes, a lack of subnational analysis, reliance on historical data, and high costs. It is further affected by recall bias, duplication, and migration of household members, and often reports pregnancy-related rather than strictly defined maternal deaths, failing to distinguish incidental cases [1,3,4]. Civil Registration and Vital Statistics (CRVS) and Health Management Information Systems (HMIS) also face challenges in accurately capturing maternal deaths. To address these gaps, the Maternal Death Surveillance and Response (MDSR) was introduced in 2004, guided by World Health Organisation (WHO) recommendations, to strengthen routine surveillance, link data to response, and provide insights “beyond the numbers” [5]. Evidence from Tanzania shows that the MDSR system has improved notification and reporting processes, but concerns remain regarding data quality, response effectiveness, and accuracy [6,7].

Despite the promise of MDSR, its performance as a reliable source of maternal mortality data has not been systematically evaluated. Existing reports highlight weaknesses in implementation, variability in the timeliness of notifications, and limited follow-through on action plans [8,9,10]. Moreover, while MDSR may yield more representative data than surveys, its validity and utility for estimating maternal mortality remain uncertain.

Therefore, this study addresses these gaps by conducting a mixed-methods assessment of MDSR implementation in Tanzania. By combining quantitative surveillance data with qualitative insights and applying NPT as a guiding framework, the study evaluates both system performance and data quality. The findings provide evidence on how MDSR functions in practice, highlight strengths and weaknesses across health system levels, and assess its potential contribution to SDG 3 tracking and accountability.

2. Methodology

2.1. Study Design and Approach

This study employed a mixed descriptive cross-sectional study that included data from January 2022 to December 2023. These qualitative methods allowed the researchers to explore perceptions, experiences, and challenges that numbers alone could not capture.

2.2. Target and Study Population

The MDSR system was the unit of assessment in this study, and data from 11 regions were analysed. In addition, structured interviews were conducted in two regions; the study population comprised regional, district, and health facility management teams, as well as healthcare providers caring for pregnant mothers and adolescents.

2.3. Sampling Technique

This study sampled all health zones as strata and regions as clusters. The zone was the centre of the sample because of its heterogeneous population, whereas regions were assumed to be homogeneous.

The following 12 regions, representing almost 50% of the country, were randomly selected: Arusha (northern zone), Manyara and Singida (central zone), Dar Es Salaam (Dar zone), Morogoro (Eastern zone), Lindi and Mtwara (Southern zone), Iringa (Southern Highland zone), Songwe (South-West zone), Katavi (Western zone), Mwanza (Lake East zone), and Kagera (Lake West zone).

Four of the twelve regions were the subject of a study to assess the success of MDSR system implementation: Arusha (Northern zone), Morogoro (Eastern zone), Singida (Central zone), and Iringa (Southern Highland zone). Two councils from each region were randomly selected, and one district council hospital and one health centre from the target region were randomly selected. In summary, four regions, eight districts, four regional hospitals, four district hospitals, and four health centres were selected.

Ten of the twelve regions were the subjects of a study of data quality dimensions, including timeliness of notification, completeness, consistency, and construct validity of the MD data, while two regions, Mtwara (Southern Zone) and Singida (Central Zone), were studied to explore perspectives on implementation of the MDSR system.

Systematic random sampling in 4-month intervals (April, August, and December in 2022 and 2023) was used to assess MDSR surveillance attributes, specifically timeliness, completeness, and consistency.

2.4. Data Collection

Quantitative data on the performance of the MDSR system were collected from four regions: Arusha, Morogoro, Singida, and Iringa. The study team adapted and used an MDSR implementation assessment tool (see Supplementary Materials) previously developed and used in Tanzania’s Lake Zone [11]. This tool tracks the pathway from daily MD notifications and weekly reporting via WhatsApp and email to MDRs, action plan development, and response. This pathway was assessed using 15 variables (

Table 1. Variables used to assess operationalisation of the MDSR system.

Dependent Variable	Independent Variable	Measurement	Description
MDSR system performance	Presence of guidelines and forms. Reports. Data display. Regular meetings. Presence of a focal person. Whether a committee appointment letter has been served. Mandatory member attendance. Regular meeting attendance by leaders. Presence of preceding minutes. Quality of the proceedings’ minutes. Trained committee. Regular supervision. Actionable recommendations. Whether implementation is discussed in the meeting. Regular feedback to the community on maternal deaths.	Overall performance of frequencies of a given variable at all levels (health facilities, district council and regional level). Overall performance of all variables at the health facility, district council, and regional levels.	Determines the frequency of performance of each variable.
Timeliness—whether the recorded value is up to date	Daily notification of incidence of or zero MD	Number of days in a year with reports of notification divided by 365 days and multiplied by 100.	Assesses each region’s daily notification rate, regardless of the presence of MDs, for all 365 days of the year.
Completeness—whether all relevant data is recorded and there is no missing data	Date of death, age, gravidity, parity, place of death, ICD 10 classification.	Proportion of MDs with a given variable, i.e., how many MDs have a date of death, age, gravidity, etc. Expressed as a % of all MDs.	Assesses the reporting rate of each of the 11 variables for all MDs per region and country.
Consistency—data values do not change in different records or systems of record keeping	Number of MDs reported	% deaths that are reported against those reported after one week	Assesses whether the total number notified within 48 h is the same as the number of MDs verified after one week.
Construct validity—the extent to which a test or measurement accurately reflects the theoretical concept or construct it is intended to measure	Age, Gravidity, Parity, ICD 10 classification, Antepartum, Intrapartum, Postpartum	% of deaths meeting ICD-10 classification with any of the variables listed.	Assesses whether each of the reported pregnancy-related deaths was a genuine MD as per the WHO definition
	Demography data: age, gravidity and parity	Age distribution of the deceased by age groups, gravidity and parity.	Assesses the distribution of MDs by age, gravidity and parity in the VMDR cases.

), and information on suspected MDs was abstracted from daily and weekly notifications.

The MS Excel 365 abstraction tool was used to extract data from the MDSR database to evaluate its quality. MDSR information, collected from daily MD notifications and weekly MD reports and transmitted via WhatsApp and email to the MOH, was used in accordance with the 2019 MPDSR guidelines.

For the qualitative component, Key Informant Interviews (KIIs) and In-Depth Interviews (IDIs) were conducted in Mtwara and Singida regions with regional, district, and facility management teams, as well as the healthcare providers responsible for maternal and adolescent care. KIIs and IDIs were used to explore perceptions, experiences, and challenges and provide contextual insights from those directly involved in the MDSR system. Trained research assistants carried out the KIIs and IDIs. A total of 38 interviews were conducted. They were recorded in Swahili, transcribed, coded, verified for integrity and anonymity, and later analysed in English. Data saturation was achieved when the research team observed recurring patterns and no new insights emerged in the later stages of data collection.

2.5. Conceptual Framework and Organisation

In this study, the implementation of the MDSR system in Tanzania was assessed. Normalisation Process Theory (NPT), developed and published in 2009 [12,13], provides a framework for understanding the drivers essential to the acquisition of new practices or implementations, for example, the MDSR system. In this study, NPT was used to assess the MDSR system by conceptualising this practice that has existed for two decades. This was achieved by examining four major categories of the NPT: (i) coherence—how people make sense out of new practice, (ii) cognitive participation—how people are enrolled and invested in work, (iii) collective action—the integration of work into practice, and (iv) reflexive monitoring—the value of given work [12,13,14].

2.6. Analysis

2.6.1. Quantitative Analysis

MDSR implementation: Microsoft Excel 365 was used to analyse the processes for MDSR implementation in health facilities, including district and regional hospitals. This analysis involved assessing 15 key MDSR variables (Table 1).

Data quality dimensions: MDSR data across 10 regions and over 24 months (2022 to 2023) was assessed for timeliness of notifications, completeness, consistency, and surveillance system attributes (Table 1).

The notification rate was assessed as the proportion of daily MD incident reports, including zero reports, from the health facility or the community, while regional notification was measured as the percentage of days with a report relative to the total number of calendar days in 2022 and 2023. The numerator was the number of daily notifications, and the denominator was 365, the number of days in a calendar year (Table 1).

Completeness was assessed by determining whether all six required data elements were completed for both the MD notifications and the weekly reports (Table 1). The expectation was that for each measured variable, there would be no missing data in every death notification, but that there would be missing data in the one verified at the end of the week. The frequency of reported variables was calculated as the percentage of MD cases in which each variable was reported.

The number of MD cases notified within 24 h was compared with the number verified after a week to determine consistency. The expectation was that the number of MDs from notifications and weekly reporting would be the same (consistent) and not vary (inconsistent). The consistency is presented as a proportion.

Construct validity was assessed by determining the number of cases that met the criteria for being defined as an MD per ICD-10 [15]. Three criteria were used: age at death (10–50 years), history of pregnancy (gravidity and parity), and/or cause of death as per the ICD classification (Table 1).

The percentage agreement was used to review the frequency of MDSR implementation, data-flow timeliness, and data-quality dimensions and consistency using an adapted intercoder reliability technique [16,17,18,19,20]. These percentage agreements were used in quantitative data analysis to map them into ordinal intervals defined as Excellent (>95%), Good (85–94%), Satisfactory (75–84%), or Poor (<75%) based on national HMIS thresholds and Kappa rankings [21,22,23].

2.6.2. Attributes of the MDSR System

The results were used to evaluate overall data quality attributes, including usefulness, representativeness, validity, simplicity, flexibility, acceptability, and reliability, as shown in

Table 2. Attributes of the surveillance system.

SN	Attribute	Explanation
1.	Usefulness	This determines whether the collected surveillance data can inform public health, including policy and performance measures such as health indicators used in needs assessments and accountability systems.
2.	Representativeness	This determines whether a representative public health surveillance system accurately describes the occurrence of a health-related event over time and its distribution in the population by place and person.
3.	Validity	This refers to the ability to capture the ‘true value’ of disease burden measures, such as incidence or prevalence, which is useful for analysing surveillance data. Internal validity concerns the extent of errors within the system, for example, coding errors when translating from one level to the next. External validity concerns whether the information recorded about the cases is accurate.
4.	Simplicity	This refers to both structure and ease of operation; the surveillance systems need to be as simple as possible while still meeting their objectives.
5.	Flexibility	This refers to the system’s ability to adapt to changes in information needs or operating conditions with little additional time, personnel, or allocated funds, including accommodating new health-related events, changes in case definitions, technology, or digital approaches, and variations in funding or reporting sources.
6.	Acceptability	Acceptability reflects the willingness of persons and organisations to participate in the surveillance system, which can be measured indirectly by the completeness of report forms and the timeliness of data reporting.
7.	Stability, reliability and adequacy	Stability refers to reliability, the ability of a surveillance system to collect, manage, and provide data properly without failure. Adequacy refers to the surveillance system’s ability to meet its objectives.

.

2.6.3. Qualitative Analysis

The analysis involved four experts, who held debriefing sessions to exchange ideas and interpretations. Disagreements on how to classify or interpret a statement were resolved through discussion until consensus was reached, ensuring reliability and validity by incorporating multiple perspectives and avoiding unilateral coding decisions [24].

The first step of analysis was to categorise participants’ group-related statements. The categories were derived from the experiences and thoughts of MDSR system users and were coded as responses. Broader categories developed research themes that explained particular phenomena [24].

In the analysis, selected quotes were used to reflect the participants’ authenticity by ensuring multiple perspectives and values were captured to promote variation among participants, using methods like ensuring the quote (i) represented a crucial point in the study from the available data, (ii) was reasonably succinct, and (iii) represented a pattern of data [24].

The convergence was used to compare, validate, and explain the quantitative results using qualitative insights and perceptions to provide an understanding of the MDSR system’s performance [25,26,27,28]. The inclusion of qualitative methods allowed the researchers to explore perceptions, experiences, and challenges that numbers alone could not capture. The goal was to provide the reader with a broader view of MDSR systems, including implementation, data quality, perceptions, and innovations.

Related codes were grouped into broader categories and then organised into themes that explained phenomena such as leadership coherence, cognitive participation, collective action, and reflexive monitoring. The themes were aligned with the Normalisation Process Theory (NPT) framework, which structured the interpretation of how MDSR practices were understood, adopted, and integrated.

2.7. Ethical Clearance

MD data collection and review are part of routine quality improvement efforts. These data are processed and stored at the MOH, Department of Reproductive, Maternal, and Child Health Services. They do not include names of the deceased and are stored and accessed confidentially; hence, they require no individual clearance.

3. Results

The MDSR system, including its implementation and data-quality dimensions, was analysed. Overall, the MDSR implementation was Satisfactory at the regional level (80%) and Poor at the district (74%) and health facility (73%) levels. While MD reporting and reviews were Excellent (95–100%) across the three assessed levels of the health system, it was observed that the response component was generally Poor.

3.1. Coherent Leadership in the MDSR System’s Success

Leadership plays a critical role in a successful MDSR system in any setting. In line with NPT, the coherence of health managers was assessed as their ability to coordinate MDSR system implementation across their councils and regions. The assessment focused on eight items: the presence of guidelines; the holding of regular meetings; member attendance; manager attendance at regional, district council, and health facility meetings; and the presence of members of MDR committees with letters of appointment. These indicators were rated Good to Excellent (85–100%) at the regional, district council, and health facility levels (

Table 3. MDSR Performance status of key implementation variables.

Variables	Levels
	Regional (n = 4)	Council (n = 8)	Facility (n = 12)	Total (N = 24)
	No (%)	No (%)	No (%)	No (%)
MDSR guidelines and forms	4 (100)	8 (100)	12 (100)	24 (100)
MDSR reports	4 (100)	8 (100)	11 (91.6)	23 (95.8)
MDSR data display	2 (50)	2 (25)	4 (33)	11 (45.8)
Regular MDSR meetings	4 (100)	8 (100)	10 (83)	22 (91.6)
MDSR focal person	4 (100)	8 (100)	12 (100)	24 (100)
Letter of MDSR committee appointment	3 (75)	7 (87.5)	11 (91.6)	21 (87.5)
Members Attendance mandatory	4 (100)	8 (100)	11 (91.6)	23 (95.8)
RMO/DMO/MOIc regularly attend meetings	4 (100)	8 (100)	11 (91.6)	23 (95.8)
MDSR meeting minutes	4 (100)	7 (87.5)	10 (83)	20 (83.3)
Quality of meeting minutes as per guidelines	4 (100)	4 (50)	6 (50)	14 (58)
MDSR teams received training	3 (75)	6 (75)	8 (66.7)	17 (70.1)
Regular MDSR supervision	0 (0)	1 (12.5)	4 (33.3)	5 (21)
Recommendations are actionable	3 (75)	7 (87.5)	11 (91.6)	21 (87.5)
Implementation discussed in the subsequent meeting	3 (75)	5 (62.5)	8 (66.7)	16 (66.7)
Regular feedback to the community on MDs	2 (50)	2 (25)	3 (25)	7 (29)
Overall score	80%	74%	73%	75%

). On this, a nurse midwife at district hospital noted that:

R1. DN. NOS1. As a nursing officer, my duties include ensuring patients are appropriately cared for; collecting and, if necessary, accounting for data; teaching and supervising subordinates; evaluating the performance and ethics of all nurses in the hospital; and, lastly, maintaining the safety of hospital property.

Coherent leadership is also evident in the daily responsibilities of leaders, who ensure the smooth provision of RMNCAH services to prevent maternal and perinatal deaths. Their primary roles include ensuring readiness for service delivery and evaluating performance, including ensuring the safety of individual pregnant women in the community. As part of efforts to prevent maternal mortality, it was found that leadership coherence extended beyond the MD reviews. The respondent, district reproductive and child health coordinator, stated:

R1. DN. NOS2. One time, I got a call that a pregnant woman at term had visited a local healer, so I acted swiftly and went to see her with a policeman. On arrival, I found a relaxed pregnant woman; we took her, and 3 days later she delivered at a hospital without complications.

Review, documentation and dissemination of findings at the facility, council and regional levels.

The regional MDSR performance reached 80%, with 8 out of 15 variables rated Excellent and 4 rated Satisfactory. The overall score for district councils was 74%, with 9 out of 15 variables rated Good to Excellent, and for health facilities, it was 73%, with 6 out of 15 variables rated Good to Excellent.

Six items for MDSR implementation were rated Poor (below 75%) at both the regional, district council, and health facility levels: meeting minutes, team training, regular supervision, discussion of implementation status in subsequent meetings, and regular feedback to the community on MDs. The poor performance of district councils was attributed to an overwhelming workload, which often led them to skip monthly reviews, and secondly, their support for health facilities, which may be regarded as a completed guideline obligation. The district reproductive and child health coordinator noted:

R2. DMK. NOS2. Council-level MDRs happen monthly; some months are skipped due to overlap with supervision. However, catch-up to complete all reviews is done.

The nursing midwife officer at a district hospital further noted that:

R2. DMA. NOR1. For an MD, regardless of facility level, the HF team is invited to the council headquarters for a review. Gaps are identified, and action plans are addressed together. Finally, the review report is sent to the regional MDSR focal person.

Health Facility Notification, Review and Response

All 12 health facilities studied scored as Excellent (>95%) in MD notification, reporting, and reviews. With regard to developing action plans to address bottlenecks, regional and district hospitals were rated Excellent and Good, respectively. However, health centres were rated as Poor (<75%). Further weaknesses in the execution of the developed action plans (responses) were observed, which were scored as Poor, with only 1 of 3 action plans implemented (Figure 1). The respondent, chief medical officer, observed that:

R2. MDC. MO1. When a death occurs, I, as the district chief medical officer, must receive information about it from all areas of my district, whether it is a dispensary, a health centre, or here at the hospital. After receiving the information, I will report to the regional level in accordance with the guidelines.

3.2. Cognitive Participation and Collective Action in MDSR System Implementation

Analysis of the second component of NPT demonstrates that cognitive participation among stakeholders expanded, thereby improving the outcomes of MDSR system implementation. Cognitive participation in this study was rated Excellent (100%) for MD reporting (notification) and reviews. However, this was not the case for action plans, with the regional hospitals scoring as Excellent (98%), the district hospitals as Good (86%), and the health centres as Poor (67%). Overall, the cognitive function of the MDSR system implementation shows a limited response to the action plan at the health facility level, with a Poor (<75%) score (Figure 1). The respondent medical officer observed that:

R2. DHI. MO2. During the MDR forum, we meet people with different knowledge and skill sets. We learn a lot in these sessions; for example, last time they showed how to use the Non-Pneumatic Antishock Garment (NASG), a lifesaving device for patients in shock.

It was found that virtual MDR was excellent at enhancing cognitive participation in reviews, action plan development, and the response component. Respondents perceived that the use of virtual MDR was effective because it enabled the cognitive, multidisciplinary participation of general practitioners, skill-mix specialists, experts, and leaders, enabling them to provide real-time expert opinions and resulting in a realistic, collective action plan from reviewing and actively monitoring responses. The district medical officer noted that:

R1. TCN. MOI. Zoom MDR enables many people to contribute to a single case simultaneously, from anywhere. Online chatting connects many people.

3.3. Collective Action and Challenges to Integrating MDSR Practice

The MDSR system in Tanzania has proven to be a powerful tool for improving the quality of care. This can be demonstrated by the high scores in notifications, reviews, and action plan development, especially in regional and district hospitals, which range from Good (85–94%) to Excellent (95–100%) (Figure 1). However, it was found that HCPs’ perceptions, views, and acceptability of MDSR implementation were negatively affected. Healthcare providers feel threatened by the process, thereby reducing their willingness to work in maternity care and affecting collective action towards developing and executing action plans, which is important for ensuring the MDSR system is integrated into practice. The respondent district reproductive and child health coordinator noted that:

R2. DN. NOS2. Honestly, HCPs have a growing fear of MDs. If we ask healthcare providers who wish to work in the labour ward, they will all probably say no.

It was also found that, during MDSR implementation, especially in reviews, the principle of “no blame, no shame, and no name” was difficult to uphold. The lack of training on how to run the MDSR system may be a reason for this. Regarding training, the findings reveal generally low scores at the regional and district levels, ranked Satisfactory (75%), while health facilities were ranked as Poor (66%) (Table 3). The qualitative data further show that the use of the VMDR approach [29], although enhancing effectiveness, appears to cause more anxiety in HCPs. The evidence shows that during VMDRs, in most cases, HCPs who managed the deceased felt gentle pressure, making them uncomfortable. This was amplified by the district reproductive and child health coordinator, who noted that:

R2. DMS. NOS2. The biggest issue with Zoom MDR is the inherent fear of the team involved in managing a deceased patient. Without running away from this reality, healthcare providers have developed a fear of the unknown.

However, other users of the MDSR system had different views on the principle of “no shame, no blame, and no name”, particularly in reviews. They felt that applying some form of gradual pressure might help reviews be more meaningful. On the other hand, the reproductive and child health coordinator further noted that:

R1. DSM. NOS2. We do indeed want a review with no shame, no names, and no blame, but if you want it to be blame-free, we likely won’t achieve it. We need to be frank about what we say. I do not want to dwell on no blame too much. What are we going to do without it? Are you recording…? (she laughed.)

3.4. Reflective Monitoring of the MDSR System

Implementation of the MDSR system has provided information beyond just numbers of MDs. In this part of the results, the main focus is on the quality of the numerical results, which may be useful for estimating maternal mortality.

Over the two years, the daily notifications (suspected MDs and zero incidences) had an interquartile range (IQR) of 155 in 2022 and 139 in 2023, and the median was 226 in 2022 and 179 in 2023. The highest reporting rate was in Songwe (83%) and Manyara (76%), rated Satisfactory (75–84%), while the lowest was in Dar es Salaam (22%) and Mwanza (28%), rated Poor (Figure 2). The overall regional average MD notification rate was Poor, at 56.4% and 43.8% in 2022 and 2023, respectively, indicating a significant discrepancy between notifications and weekly reporting (Table 2).

Interviews indicated that the DRCHco oversees daily notifications from health facilities and communities at the DMO offices. They receive information regarding sick mothers and children, including deaths, through calls, SMS, and WhatsApp groups. The respondent district reproductive and child health coordinator observed that:

R1. DM. NOS2. I receive daily reports on maternity outcomes. Like, one facility in my district council had an MD who immediately called to say things are not OK on my side!

MD notification data showed internal inconsistencies in completeness, ranging from 73% for gravidity to 91% for place of death, with Songwe, Iringa, and Manyara scoring as Excellent and Arusha generally scoring as Poor. There was a significant improvement in data completeness for weekly reporting, with all six variables scoring as Excellent (95–100%), except for assigning cause of death (ICD-10), which was Satisfactory (82%). Similarly, the findings showed inconsistencies in accounting for MDs, with fewer notified (200) than reported weekly (349) (

Table 4. Weekly maternal death report completeness (N = 349).

Region		No. of MDs	Date of death	Gravidity	Parity	Age (10–50 Years)	Place of Death	ICD 10	Average
Arusha	Notify	14	57%	29%	29%	57%	86%	-	52%
	Weekly	27	100%	100%	100%	100%	100%	100%	100%
Dar Es Salaam	Notify	26	100%	15%	100%	100%	100%	-	83%
	Weekly	95	100%	66%	100%	100%	100%	46%	93%
Iringa	Notify	11	100%	91%	100%	82%	100%	-	95%
	Weekly	13	100%	92%	100%	100%	100%	77%	98%
Kagera	Notify	11	100%	91%	82%	100%	91%	-	93%
	Weekly	16	100%	100%	100%	100%	100%	88%	100%
Katavi	Notify	13	62%	54%	54%	62%	77%	-	62%
	Weekly	18	100%	100%	100%	100%	100%	89%	100%
Lindi	Notify	16	75%	88%	88%	88%	81%	-	84%
	Weekly	16	100%	100%	100%	100%	100%	75%	100%
Manyara	Notify	20	95%	95%	100%	95%	90%	-	95%
	Weekly	22	100%	100%	100%	100%	100%	73%	100%
Morogoro	Notify	42	93%	93%	93%	90%	93%		92%
	Weekly	57	100%	98%	98%	100%	100%	75%	99%
Mwanza	Notify	33	88%	85%	88%	82%	88%	-	86%
	Weekly	72	100%	100%	100%	100%	100%	100%	100%
Songwe	Notify	14	100%	86%	93%	100%	100%	-	96%
	Weekly	13	100%	100%	100%	100%	100%	92%	100%
Average	Notify	200	87%	73%	83%	86%	91%	-	84%
	Weekly	349	100%	96%	100%	100%	100%	82%	99%

Notify—daily notification, Weekly—weekly reporting, - Not assigned during notification.

).

3.5. Attributes of Data Quality: Usefulness, Representativeness, Validity, Simplicity and Flexibility of the MDSR System

In line with NPT, this study underlines the reflexive nature of the value of MDSR system implementation. Readers may understand this by examining the attributes of the MDSR surveillance system, including its usefulness, representativeness, validity, simplicity, flexibility, acceptability, and reliability, which are summarised in the

Table 5. Attributes of data quality.

SN	Attribute	Survey Reflection
1.	Usefulness	The collected MDSR data may inform an accountability framework, affording improvements in stewardship and healthcare providers’ skills, knowledge, practices, and attitudes (Table 3 and Figure 1).
2.	Representativeness	MDSR data provide MD distribution patterns by zone and region (Table 4).
3.	Validity	The use of ICD 10 for classification serves as an indicator of real MDs and pregnancy-related deaths. In this case, internal validity errors were minimised by reviewing the notified suspected MD data and conducting weekly MDRs to determine the causes of death. The ICD-10 tool ensured the external validity of MDs, aligning with the criteria (Table 4).
4.	Simplicity	The MDSR data collection process was simplified by using the WhatsApp digital platform for notifications, a designated email account for weekly verification, and conducting a daily online MDR (Figure 2, Table 4).
5.	Flexibility	The MDSR system in Tanzania appears to have adapted to technological changes, as demonstrated by online approaches that facilitate easy notifications, reviews, and plan execution (VMDR).
6.	Acceptability	The MDSR approaches are universally accepted by implementors, though there is a growing concern about upholding the principle of blame, as evidenced by improved reporting after one week of MD occurrence and improved review scores.
7.	Stability, reliability, and adequacy	The MDSR system is highly stable, having provided data for this study and others and having been operational for almost 10 years. Through this system, continuous surveillance is conducted, and strategic data are provided to inform policy, resulting in a number of publications.

below.

4. Discussion

This study provides insight into MDSR system implementation and its data generation and quality, reviewing the opportunities for it to inform the health sector beyond just numbers. The use of NPT and the convergence of quantitative and qualitative information help provide an understanding of the implementation of the MDSR system practice in Tanzania.

4.1. Effect of Coherent Leadership

It is demonstrated that coherent leadership is the single most important factor in enhancing health system productivity as it provides a political environment conducive to improving health system performance, including through the implementation of the MDSR system. The system’s super users, including stewards, implementers, and community clients, were identified, with thorough follow-up. It has been demonstrated that, through MDSR system implementation, leaders not only conduct surveillance and review MDs, but also perform follow-ups of pregnant women to prevent morbidity and death.

4.2. Cognitive Participation and Collective Action as Means to Improve Implementation

This study revealed remarkable Satisfactory and Good-to-Excellent performance in most activities related to MD surveillance and reviews across all governance levels and health facilities. As demonstrated in the virtual MDR, cognitive participation is a strong catalyst for better reviews, as it brings together different experts with different skills to enable a meaningful discussion [29]. A remarkable improvement in the development of action plans at regional referral hospitals and district hospitals was observed, but not in health centres. This may be attributed to the presence of highly knowledgeable and skilled healthcare providers at higher levels within the health facility hierarchy, which is not present at lower levels such as in health centres. This finding is in agreement with those of three studies [9,29,30,31], which noted that a successful MDSR system requires functional cognitive participation of knowledgeable and skilled healthcare providers.

4.3. Collective Action Is Affected by Blame Culture

The findings demonstrate that the MDSR system is an effective tool for improving care by providing both numerical and qualitative data. However, as found in many studies, blame and fear remain major barriers to implementation, hence limiting the scope of collective participation. Evidence from eight peer-reviewed articles shows that blaming and harsh language tend to inhibit participation, attendance, and staff commitment [11,32,33,34,35,36,37,38]. The challenge ahead pertains to crafting an MDR meeting guide and maintaining coherent leadership during reviews, which would prevent finger-pointing at any person or health facility.

Kinney et al. (2020) published 10 strategies to address fear and blame when implementing an MDSR system [37]. These strategies include policy and planning, a national policy to end maternal and neonatal deaths, harmonisation of guidelines into a standardised tool, the creation of and support for an enabling environment for MDSR system implementation, strong leadership, teamwork, regular multidisciplinary reviews, a code of conduct during reviews, respect for individual roles, and promotion of community engagement and awareness [37,39,40].

The strategies above are essential for successful MDSR implementation. The virtual MDR was launched in Tanzania in 2021, alongside conventional reviews, to mitigate fear and blame stemming from broader participation and the views of high-level leaders. To control ego amid this diverse group and enhance collective participation, a decision was made to prepare guidance for virtual review meetings [29]. The MDSR study conducted in Suriname (2021) proposes ways to improve MDSR implementation by taking into account the following issues: commitment, ‘no blame, no shame’ culture, coordination, collaboration, and communication [40].

4.4. The MDSR System Enhanced Reflexive Monitoring

This study has shown that notifications are not timely, i.e., within 24 h, as per standard practice. The leading model in this study, normalisation of MDSR practice, requires that the system be incorporated into the electronic patient care system. Currently, reliance on WhatsApp and paper-based tools does not meet the requirements for Excellent reporting. Improved timeliness would directly affect notification rates, potentially increasing the current MD data yield from 57.3% (Poor) to nearly 100% (Excellent). This may enhance the reliability and consistency of the MDSR system as a source of routine MD data.

Published reports show that data completeness is a challenge in MDSR systems [41,42]. This challenge has been averted in Tanzania by tracking data through notifications and weekly reporting, rated Poor and Excellent, respectively, achieving 95% coverage. This finding sheds more light on reporting low- to middle- income countries like ours, where routine data remains a more reliable source of MD data than CRVSs [43]. It was shown in this study that a combination of notifications and weekly reporting increased the reporting of MDSR data, which may be used to estimate maternal mortality.

4.5. Organisational Barriers Related to MDSR Implementation

The study revealed weaknesses in organisational processes, including poor documentation of meeting minutes, irregular supervision, and skipped monthly reviews due to overlapping responsibilities. Feedback to communities on maternal deaths was limited, with only 29% of facilities providing updates, which undermined accountability and learning at the grassroots level.

4.6. Financial Barriers Related to MDSR Implementation

Although not directly quantified, financial constraints were evident in the reliance on low-cost communication channels such as SMS, WhatsApp, and email for reporting. District councils faced overwhelming workloads and resource shortages, suggesting inadequate funding to support regular supervision, training, and more robust digital health infrastructure.

4.7. Human Resource Barriers Related to MDSR Implementation

Human resource challenges included insufficient training of MDSR teams, with only 70% having received formal preparation. Staff at district level were often overburdened, leading to skipped reviews and reduced support for health facilities. Dependence on a few focal persons and committee members further strained capacity and risked burnout.

4.8. Governance Barriers Related to MDSR Implementation

Governance gaps were evident in the poor execution of action plans, which were rated as Poor at health facility level, with only one of three implemented. While reporting and reviews were Excellent, accountability mechanisms for enforcing corrective actions were weak. Coordination between regional, district, and facility levels was fragmented, limiting the translation of recommendations into practice.

5. Limitation

The study was limited to a description of MDSR implementation to provide a comprehensive view of MDSR system performance and was unable to associate the implementation with other outcomes, as exposure-outcome data were not collected. The study used data from notification and reporting and did not use virtual MDSR. The qualitative study included only providers directly involved in the MDSR implementation.

6. Conclusions

The study highlighted that the Tanzania MDSR system has strengthened maternal death data generation, with notification and reporting rated Excellent and weekly reporting consistently achieved. These processes enabled regular reviews and actionable plans, though execution at the facility level was weaker. Qualitative findings highlighted innovations such as VMDR, which expanded participation and improved engagement. Together, these results demonstrate that MDSR data go beyond numbers to reveal systemic gaps and guide corrective actions, contributing to improved maternal care in line with SDG 2030. However, reliance on paper-based tools limited efficiency, and integration into an electronic patient management system was recommended to streamline data collection, enhance analysis, and support better health outcomes.