Maternal Death Surveillance and Response in Tanzania: Comprehensiveness of Maternal Deaths Narrative Summaries and Action Points from Death Reviews


 BackgroundReview of maternal deaths relies on comprehensive documentation of medical records that can reveal sequence of events that led to death. Maternal Death and Surveillance (MDSR) system recommends the use of narrative summaries during maternal death reviews to discuss the case and categorize medical causes of death, identify gaps in care and recommend action plans to prevent deaths. Suggested action plans are recommended to be Specific, Measurable, Attainable, Relevant and Time bound (SMART). To identify gaps in documenting information and planning recommendations, comprehensiveness of written narrative summaries and adequacy of action plans according the MDSR guideline were assessed. MethodsA total of 76 facility maternal deaths that occurred in two regions in Southern Tanzania in 2018 were included for analysis. We assessed the comprehensiveness of narrative summaries and action plans using a prepared checklist from MDSR guideline of 2015. Presence or absence of items in four domains each with several attributes was recorded on the checklist. The domains were socio-demographic characteristics, antenatal care, referral information and events that occurred after admission. Less than 75% completeness of attributes in all domains was considered poor while >94% was good/comprehensive. Action plans were assessed by application of SMART criteria and according to place of planned implementation (community, facility or higher level of health system). Results Two-thirds of summaries (66%) scored poor, and none were scored as good/comprehensive. Summaries missed key information such as demographic characteristics, information of events that occurred in community (16%), time between diagnosis of complication and commencing treatment (65%), investigation results (47%), summary of case evolution (51%) and referral information (47%). A total of 285 action plans were analysed. Most action plans 242(85%) were allocated to health facilities for implementation and they were mostly 42(42%) on service delivery. Only 42% (32/76) of the action plans were deemed to be SMART.ConclusionsAbstraction of information to prepare narrative summaries used in MDSR system is inadequately done. Action plans and recommendations in MDSR system are mostly for facility sub standards of care and are not specific on the issues to be addressed.

countries including Tanzania, to reveal the causes and contributing factors to maternal deaths, with some success and challenges (2)(3)(4)(5)(6)(7). In 2015, Tanzania introduced the Maternal Death Surveillance and Response (MDSR) system in line with World Health Organization (WHO) recommendations (8,9). It is one way to address high Maternal Mortality Ratio (MMR), by uncovering local solutions for local problems and to guide national strategies towards improving quality of care. The MDSR system includes identifying, notifying and reviewing all maternal deaths to describe: (a) medical causes of deaths, (b) shortcomings/delays in the health system that contributed to the death, and (c) recommendations to address the identi ed delays. The recommendations address the underlying medical cause of death and delays in care identi ed from the community to health system.
Facilities providing childbirth service in Tanzania are sought to have a multi-disciplinary MDSR committee to review and audit maternal deaths. The MDSR guideline includes instructions and illustration on the collection of information from i) medical les, ii) interview of health care providers and iii) interviews with relatives who cared for the woman before death (8). The information is used to prepare a narrative summary for discussion during MDSR meetings at health facility, district and sometimes regional level of the health system. More information is sought in medical les (when available) during the meetings if what is written in the summary is not su cient. The summaries are kept as con dential documents in hard or electronic copies by a designated person in the facility. After the meeting one or more recommended action points are suggested by the committee and lled in the maternal death reporting form. The action points are meant to stir up response at local and national level to prevent future deaths (10). The recommended actions from maternal reviews need to have clearly de ned and measurable activities so that implementation can be tracked and assessed. That means they ought to be Speci c, Measurable, Attainable, Relevant and with speci c allocated Time (SMART) and responsible person for implementation. (See Fig. 1. The action points are then shared to the district health o ce and quality improvement committees for further follow up. Having a system of following up quality and implementation of recommended action points can be effective in making sure MDSR is implementable. This has been shown to be effective in other countries which implement MDSR. A study from Nigeria reported use of scorecard to track MDSR implementation pointed out facilities with recommended actions without clearly de ned activities (10). This could have created problems during implementation and follow up. The MDSR guideline of 2015 in Tanzania does not provide a framework for follow up of implementation of action plans but recommends them to be SMART.
Comprehensive documentation of history, physical examination, investigation results and treatment in medical practice is important in assisting practitioners and other medical staff to manage, follow up patients, used the information in research and audits/reviews to improve practice and patients' safety (11)(12)(13). During death reviews such as those in MDSR system, the quality of documents used may directly impact the recommendations from the audit, especially when it is done from abstracted information. Oftentimes, health care providers fail to follow guidelines in documenting patients' information by prioritizing care over documentation and so the abstracted information may be inadequate (14). Studies from United States and Iran have shown record keeping in medical les in health facilities are weak and face challenges as health care providers fail to follow recommendation during gathering and storing of information (12,(15)(16)(17). While in Northern Tanzania, a report on MDSR implementation by Maternal and Child Survival program and other partners revealed most facilitiesm edical records were not su cient to decide the cause of death and substandard care (18).
In view of poor record keeping and documentation of medical les in health facilities in Tanzania, the maternal deaths narrative summaries need to be comprehensive since there is a chance to gather information from multiple sources. Therefore it is imperative to reveal the shortcoming in writing and storing information in the summaries inorder to recommend way forward.
We sought to investigate the availability and comprehensiveness of the summaries in health facilities, and assess how well action plans aligned with the SMART criteria. Results will provide recommendations for improvement of record keeping and gathering of information in the maternal death narrative summaries.

Study design
This was a retrospective desk review of maternal deaths documents (narrative summaries and action plans). To do this we visited all facility that reported deaths between 1 st of January to 31 st of December 2018 for Mtwara and Lindi regions of Southern Tanzania and reviewed the narrative summaries and action plans. A total of 122 maternal deaths were followed up in the facilities for their summaries and action plans

Study setting
The total population of Lindi and Mtwara regions is about two million people (19) . There are two regional referral hospitals, 12 district hospitals, four private/mission hospitals, 40 health centres and 399 dispensaries. In 2015, the MMR was 456 in Lindi and 579 per 100,000 live births in Mtwara (20). Facility delivery was 80.8% and 81.3%, caesarean section rate 6.0% and 10.3% respectively and family planning use was at least 50% in both regions (21). The two regions like all other regions in Tanzania have an MDSR system through which all maternal deaths occurring in health facilities are routinely reviewed. Maternal deaths occurring in each facility are reported to the district and ultimately to the region and Ministry of Health, Community, Development, Gender, Elderly and Children.

Outcomes
We reviewed the deaths documents following a de ned set of criteria. We de ned Comprehensiveness of narrative summaries as summaries that have more than 94% of the information that is recommended by Tanzania MDSR guideline of 2015. The information in the summaries was divided into four domains each with several attributes (Panel 1).

Data sources and measurements
A team of researchers led by the rst author (AS) visited the health facilities in March and April of 2019 and requested the narrative summaries of all 122 noti ed maternal deaths from the facility in-charge. The rst author (AS) reviewed the narrative summary using a checklist informed by recommendation in the 2015 Tanzania MDSR guideline (8).
The narrative summaries were assessed by familiarisation and checking for presence of attributes on the four different domains (Panel 1). Presence or absence of information/attributes in each domain was scored and coded as present (1), not present (2) or not applicable (3) depending on the case. The researcher read each summary repeatedly to make sure all information was available or not even if it was not explicitly mentioned. For example, the duration of amenorrhea was considered to be present if the last normal menstrual period was mentioned even if the gestation age was not mentioned explicitly. Also, marital status was considered to be present if it was mentioned that the deceased was brought to facility by husband.
After familiarisation with the action plans the rst authors extracted i) the target of each action plan (community, facility or higher level) ii) speci c issues it addressed in the community or facility. For community action plans, the researcher indicated whether the action was for decision making at family level, danger signs recognition or health seeking behaviour or traditional practices. Action plans in the health facility were assessed whether they addressed service delivery, human resource, equipment and supplies, referral system, accountability or facility infrastructure. The action plans were then assessed for appropriateness by checking whether they met the SMART criteria.

Quantitative variables
Quantitative data collected was entered and cleaned in SPSS version 23 for analysis. The Comprehensiveness of each narrative summary was determined by calculating individual proportion of amount of information depending on each case. We summarised the total amount of information for each summary and then the proportion of present (1) was calculated from the expected total score for that case. The proportional score of each summary was ultimately divided into be poor, average, or good/comprehensive if it had 0-74%, 75-94% or more than 94% of the required information respectively. The cut off points were decided based on having been used in a study done by Mohseni et al in Iran (15), and were used for analysis and description purposes and are not recommended as standard cut off levels. Action plans were considered to be SMART if all the criteria were met.

Statistical methods
Descriptive analysis was done for all variables and data presented in gures and tables.

Assessment Of Comprehensiveness Of Narrative Summaries
Each narrative summary is recommended to have demographic, antenatal care information, delivery information (if delivered before admission), referral information (for referred cases), and information of events after admission until death. Age and gravidity were the most common information 69(91%) present in the summaries, while only 7(9%) of the summaries had a maternal death review number indicated. Only 8(13%) summaries indicated the mode of delivery of previous pregnancy and only one had a date of previous caesarean section. (Table 1) The table also indicates that for those who delivered before last admission, 18(95%) summaries had information on date and place of delivery/abortion while 8(42%) had information on the duration of amenorrhea.
Most summaries 28(88%) indicated the type of referring facility, while none of them indicated "how the woman`s position in the community affects her referral" as recommended in the guideline  (Table 4). Information on how the woman`s position affects the process after admission was not present in any of the summaries. Overall, 64(84%) of summaries were scored to be poor and only 12(16%) were average and none were good/comprehensive. When the two variables that scored zero were removed (Tables 1 and 2) the summaries scored changed to 66% poor and 34% average.

Assessment Of Recommended Action Points After Maternal Death Reviews
A total of 285 action plans were included in the analysis. Of the reviewed action plans, 242(85%) included recommendations targeting the facility, 42(15%) the community and 0.4% higher level of health system. Almost half 120(42%) of the recommendations directed to the facility were for service delivery, such as knowledge and skills, while in community most action plans were for delays in decision making (Fig. 2).

Recommended action points assessments
Two summaries did not have documented action points. A total of 285 recommended action points were included in the analysis  Table 5 shows that approximately 42% of the action points were SMART, and most of them were time bound (93%) and were deemed relevant (71%).

Discussions Main ndings
Our study reveals that only 62% of narrative summaries for maternal deaths were available and none had all the recommended information as according to 2015 MDSR guideline. Missing key information included information of events that occurred in the community before reaching facility (16%), time between diagnosis of complication and commencing treatment (65%), investigation results (47%) and summary of case evolution after complications (vitals, input, output, treatments given) (51%). Furthermore, just over half of referral deaths had summary of the medical history, physical examination and treatment of case before referral (53%). Demographic characteristics such as death review number, patient code, and marital status, duration of amenorrhea and mode of delivery of previous pregnancy were missing in most summaries. Most action points (85%) were directed to health facilities and they were mostly targeting service delivery issues such as knowledge and skills due to human error in management. Only 42% of the action plans were deemed to be SMART, most of the action plans (93%) had time line of implementation while less than half (46%) were found to be speci c.

Comprehensiveness of the narrative summaries and action plans
This study con rms that MDSR systems are constrained by poorly prepared narrative summaries lacking important information. Studies in the US and Wales, UK have shown that medical les have poorly documented general symptoms, gynaecological history, treatment side effects, smoking history and drug allergies (16,22,23). This has direct negative impact on the comprehensiveness of summaries abstracted from such documentation. Luck et al in a study on quality of abstracted information in general internal medicine patients, cautioned against measuring quality using abstracted information as it may have many de ciencies. They reported that chart abstraction resulted into only 54% of the standard information required (14).
One of the reasons for poorly documented narrative summaries in our study could be the fact that, a person who was involved in the management of the deceased is tasked with writing the summary. This could lead to attempt of hiding some of the information in the summary in fear of blame. A study in Malawi reveal that fear of blame was one of the main barriers to conduct of maternal deaths reviews in health facilities (24). This problem could have been mitigated by assigning a different person to prepare the summary and providing a comprehensive example narrative summary in the guideline (8). This should serve as a wakeup call for facilities to appoint a single designated person to write the narrative summaries and a guideline to have a friendlier user guide for preparing summary.
Documented recommendations or action plans in the MDSR systems were mostly directed to health facility (third delay) targeting directly health care provision such as knowledge and skills of health care providers. This seems to be reasonable as also other studies from Malawi, Tanzania, Kenya and Nigeria indicated that most maternal deaths occur due to substandard care in health facilities (25)(26)(27)(28)(29) while facility delivery in these countries stands at 91%, 63%, 61% and 39% respectively (21,(30)(31)(32).
For the action plans to be effective in preventing and reducing maternal death they need to be implementable and easy to follow up. Most action plans in the MDSR system were found to be nonspeci c (54%) as they were not clear about what was going to be done and only 42% were found to be SMART. This may limit the impact of the MDSR strategy on quality of care in Tanzania. Few studies have assessed the recommended action plans MDSR system such as in Nigeria and in Northern Tanzania (10,18). During maternal deaths review, health care providers should put down recommendation with implementation in mind.

Limitations of the study
The main limitation of this study is the fact that the summaries were assessed by one person (AS). Bias was however minimized by using a prepared checklist from MDSR guideline recommendations. The cutoff points used in analysis of information were informed by a similar study from Iran (15) but may be a matter of discussion. In our study we used them for description purposes only and do not indicate standards in amount of medical information documented.
The generalisability of the study in settings where MDSR system does not use narrative summaries is also a limitation. Even in these settings, the study informs the importance of measuring quality of care using comprehensive medical information. It also shows the effects of incomplete documentation of medical information as it affects the quality of abstracted information.

Conclusions
Abstraction of information to prepare narrative summaries used in MDSR system is inadequately done.
This can impact negatively the quality of care measured using the summaries. Action plans and recommendations in MDSR system are mostly for facility sub standards of care and are not speci c on the issues to be addressed.

Recommendations
To improve documentation of narrative summaries and recommended action points, providers should use a checklist with spaces to ll the required information. A scheduled follow up of action point implementation is needed to ensure reviews work as intended  Figure 1 The process of maternal death review by MDSR committee