Study setting

We focused on public-sector services; Bangladesh and Tanzania have national strategies to improve maternal and newborn health (MNH) data [27, 28]. Both countries have well-established RHIS, including the national use of District Health Information Software-2 (DHIS-2) for collecting, managing, and using electronic RHIS data [29]. In Bangladesh, public-sector MNH services are provided by the Director-General Health Services (DGHS) and the Director-General Family Planning (DGFP) [30]. In Tanzania, public MNH services are governed by the Ministry of Health and Social Welfare [31].

We embedded this qualitative research within a mixed methods study that developed and tested tools to improve the availability, quality, and use of newborn and stillbirth indicators in RHIS; the Every Newborn-Measurement Improvement for Newborn & Stillbirth Indicators (EN-MINI)

Tools [32]. Although the EN-MINI tool focus was more general, we were able to undertake our qualitative data collection in the same public health facilities, plus national and regional offices at the same time.

Design

We undertook this phenomenological qualitative research within a constructivist paradigm using in-depth interviews in Bangladesh (1^st Sept-21^st Dec 2020) and Tanzania (21^st Feb- 30^th April 2023). The semi-structured interview guide was adapted from the ’Assessing Barriers to Data Demand and Use in the Health Sector’ tool [33] (S1 Text).

Mapping of data elements in registers

We mapped the data elements of the standardised paper-based L&D registers used at participating health facilities. At the time of data collection, standardised OT registers were not being used for caesarean births and were therefore excluded from mapping.

Participants, sampling, and recruitment

Participants were from facility, sub-national, and national levels of the healthcare systems in Bangladesh and Tanzania. Participants were purposively selected to include a diverse range of roles, levels of care, and responsibility at different levels of the health system health (including health and data professionals, managers/leaders, and policymakers) (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Participants by level of the health system.

Fig 1 shows 2 pyramids side by side, one to represent Bangladesh and one to represent Tanzania. The base of the pyramid represents the different health facilities and offices that participants were recruited from at community level (colour coded in blue). The next layer represents the district-level facilities and offices, these are dark green. For Tanzania only, the regional level office is in dark orange. At the top of the pyramid, national-level offices are grey.

https://doi.org/10.1371/journal.pgph.0003808.g001

In Bangladesh, district and facility-level participants were selected from both Directorates (DGFP and DGHS) in Kushtia district. The sampling frame included district-level offices (2), Upazilla-level office (1), district hospitals (2), Upazilla health complexes (5), Union sub-centres (3), community clinics (2). In addition to national-level offices (2) for DGFP and DGHS.

In Tanzania, district and facility-level participants were selected from two districts in Tanga Region: the sampling frame included district-level offices (2), hospitals (2), health centres (4), and dispensaries (4). As well as the regional and national offices (2).

Data collection and management

Experienced qualitative researchers (TM, DS, SM) interviewed participants in their preferred language (Bengali, Swahili, English). Interviews took place in a private and mutually convenient location of the participant’s choice; they lasted 45–75 minutes. Seventy-seven participants consented to audio-recorded interviews, and two (1 for each country) consented to written documentation instead. Further information concerning non-verbal communication, and the general tone and atmosphere were documented in an interview summary. Research assistants transcribed and translated the interviews into English, and transcripts were checked by the interviewing researchers.

The research team engaged in regular reflexive meetings at all stages of the study, considering our positionality, the conceptualisation of preliminary themes, and the ongoing refinement of the interview guide and recruitment strategy (S2 Text). Data collection and analysis were concurrent, allowing emerging themes to be investigated in later interviews. The sample size was met when no new information emerged about our research questions.

Transcripts were de-identified, and aggregated descriptors were used. The participant’s demographic characteristics are stored separately from the transcript. All records are managed and stored according to agreed data protection guidelines.

Ethical considerations

This study was granted ethical approval by the Bangladesh Research Review Committee (icddr,b), Ifakara Health Institute, the National Institute for Medical Research (Tanzania), and the London School of Hygiene & Tropical Medicine (United Kingdom) (S1 Table). Information regarding the study objectives, process, and intended data use was shared with potential participants in line with approved participant information and written consent forms (S3 Text). Participants were reassured of their right to withdraw consent without consequence. We followed the Standards for Reporting Qualitative Research recommendations (S1 Checklist).

Qualitative data analysis

HR, DS, and TM undertook thematic analysis following the six phases developed by Braun and Clarke [34, 35]: familiarisation with data, initial line-by-line coding, theme generation, review, definition and naming of themes, and finally, synthesis and conclusions. Supported by NVivo software [36], both inductive and deductive analytical approaches were used. Initial open coding informed the development of the coding framework and the data flow (Fig 2). Next, codes were organised using categories from the Performance of Routine Information System Management (PRISM) framework [37]. PRISM includes three main determinants of RHIS performance: a) Technical factors relating to the tools and technologies used within RHIS’ for data collection, processing, and analysis. b) Organisational factors relating to the policies, leadership, and culture underpinning the RHIS structure and environment. c) Behavioural factors relating to individual perceptions and experiences that interact with how RHIS users support and engage with the RHIS [38]. Consensus was reached through regular analysis discussions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Data flow by place, people, and care pathway comparing vaginal births on labour and delivery ward and caesarean sections in operating theatre.

This flow chat maps the care and data journey from admission to birth for vaginal and caesarean births. Care activities are shown in green and linked to a pregnant person icon, data activities are in grey and linked to a chart icon. The key staff responsible are also mapped using icons to represent different professionals.

https://doi.org/10.1371/journal.pgph.0003808.g002

Data flow and registers differed by mode of birth with potential RHIS data gaps for caesareans (technical factors)

In Bangladesh and Tanzania, having a caesarean changed the location of birth (L&D ward/OT), availability of health professionals, and care pathway, impacting the data flow (Fig 2). In all facilities, participants explained that admission and triage were initiated in the L&D ward. Women progressing to a vaginal birth remained in the L&D ward, and care was usually provided by a designated nurse or midwife. Those requiring a caesarean were transferred to the OT, with referral to a different facility if surgical services were not available.

In Bangladesh, both the Directorate for Health Services (DGHS) and Directorate for Family Planning (DGFP) use the Emergency Obstetric and Neonatal Care (EmONC) register (rather than an OT register) for all modes of birth. The EmONC registers are similar however, DGFP included three data elements not collected in the DGHS (last monthly period, ’reason for a caesarean’, ’type of anaesthesia’) (S2 Table). The EmONC registers were kept in the L&D wards (rather than in the OTs). Some facilities occasionally used an older version of the delivery and OT registers during stockouts of the current EmONC register. Co-ordination for L&D and OT data was challenging and health professionals generally prioritised completing medical records during the admission, leaving the register filling until discharge.

‘We don’t write all their information when they get admitted. Suppose a patient came with labour pain, in this case, until the delivery is done, we don’t fill out the [register] cells. That remains blank. When everything is done and she will be discharged, then we fill out the cells.’

Midwife, District Hospital, BD

In Tanzania, the national RHIS, Mfumo wa Taarifa za Uendeshaji Huduma za Afya (MTUHA), is a series of registers designed for primary data collection (’MTUHA 12’ for L&D ward). Standardised OT registers are not used and processes for recording caesarean birth data in the register varied between facilities. In five facilities in Tanzania, a L&D ward midwife would move to the OT to provide immediate newborn care and collect/complete the L&D register on their return. Many facilities had informal (hand-written) registers, referred to as ’counter’ books, for use in the OT. Participants reported using this information to complete and cross-check documentation (e.g. the number of caesarean births) in the L&D ward register. However, limitations were mentioned: a regional manager said, ’If you use the counter book, you won’t collect all the data because there is not enough space to be able to fill what is required.’ Two participants (district and facility-level) reported concerns that data regarding caesareans were miscounted and there were noted recording gaps regarding care provision data for caesarean births, for example: ’…details for anaesthesia such as complications or how much the patient was bleeding, we don’t have a place to record that information.’ Nurse in-charge of OT (District hospital), TZ.

Neither Bangladesh nor Tanzania had a standardised process for managing data flow for caesareans leading to potential RHIS data gaps for caesarean births.

Digitisation worsened the duplication for fragmented RHIS and increased the data burden, especially for overstretched staff on L&D wards (technical factors)

In Bangladesh and Tanzania, duplicate RHIS tasks required health professionals to enter the same information in multiple paper-based and electronic systems (e.g., individual medical records, different registers, forms, and tally sheets, or databases). This was highlighted by staff from all facilities in Bangladesh and 8/10 facilities in Tanzania. L&D staff (and staff in lower-level facilities) were more impacted by duplicative processes than those working in OT, for example:

’If a mother enters the labour ward, we have a register with 47 columns where we enter… there’s information about ANC and admission. There’s some about PNC as well. We also have to make online entries for each admission in the labour ward. ANC doesn’t have online entries, but labour ward does.’

Midwife (District hospital), BD

Implementation of electronic RHIS systems differed between and within countries, including processes for where and who digitised RHIS data. Many staff lamented that the introduction of eRHIS which required them to complete both paper-based and eRHIS records with duplicate information, and created a double recording burden.

’We usually have double work, which means we have to get time to record data on paper and then record it in the system. This increases the workload; when you are busy, it reaches a point where you will not record in either.’

Nurse, recovery (District hospital), TZ

In Bangladesh, both directorates were progressing towards scaling up DHIS-2 within all facilities. As with the paper-based registers, data for vaginal and caesarean births were aggregated together in the same DHIS-2 form but poor alignment between eRHIS forms and the registers was noted by a district manager as challenging.

Tanzania also used DHIS-2 software. In addition, several facilities used supplementary RHIS software (AFYA CARE, Kanzi Data, and GoT-HoMIS) and highlighted the imperative to integrate systems and remove redundancy.

’I think that if GoT-HoMIS is integrated it will be much better than entering data in GoT-HoMIS, and in the paperwork, at the end of the month, you enter the report manually, then enter it in the system. If they are integrated, after completing to enter data in GoT-HoMIS at the end of the month you will just generate the data, such as exporting in DHIS2. You get rid of the paperwork.’

Regional manager, TZ

These technical factors contributed to the duplicative data burden and fragmentation that electronic systems should help solve. Especially given that health professionals from the L&D ward were responsible for a large variety of different registers and forms. Many extended well beyond the scope of intrapartum care and remined largely paper-based, despite implementation of electronic systems (DHIS-2):

’There are a lot [of registers]! For example, childcare, general patient care, mother and newborn baby care, inventory, and combined register. In addition to that, there is also a pill supply register and two different registers for injections. There is also an IUD [intrauterine device] register with three different sections. I have to fill out those as well.

Senior staff nurse (District hospital), BD

There were also noted design challenges for registers. Despite the plethora of existing registers, some participants requested a separate register for newborns, especially in Bangladesh, where the register included many free text boxes. Six participants from Bangladesh (none from Tanzania) complained that the cells were too small. Three health professionals from Bangladesh also highlighted that the small cell size impeded the accurate medication recording.

Each data entry (whether paper-based, or at the point of digitisation) represents an opportunity for error and an associated requirement for quality-checking processes. Quality checks are necessary, but participants explained that this was labour-intensive and involved multiple staff at different health system levels:

’It means that once the report is written, the final monthly report must be reviewed by two or three people from the ward before we submit it.’

Assistant nurse (District hospital), TZ

’Then our in-charge and seniors come and find our mistakes. And we are accountable to these mistakes… they are checking the register book and also cross-checking the registered data to check whether we have given them the authentic information.’

Community health care provider, BD

’[During a spot check] I’ll open a certain Register, open Tally sheet, open a summary form and open DHIS2. Is the information entered in those areas the same? That is mostly what is checked. Number two, are the information entered realistic? This spot-check should be done daily, but sometimes because of people being busy, it becomes hard to do daily.’

Nurse In-charge (Dispensary), TZ

This top-down quality process had some unintended consequences which are detailed in the below section regarding organisational factors. Our findings also highlight the missed opportunities for better integration of mechanisms within electronic systems to monitor and support data quality.

Health professionals were overwhelmed by duplicative RHIS tasks which competed with providing care (behavioural factors)

Participants’ testimonies consistently focused on insufficient human resourcing more than other dimensions of RHIS performance. The sense of being overwhelmed permeated accounts from healthcare professionals in both countries, for example:

’We have to maintain maternal care, and we also have to look after the baby as well. So, there is often a rush of patients, so we stay busy. We have to clean up the mess after the delivery and also take care of the baby. I feel disoriented in those moments.’

Community Health Care Provider (Community clinic), BD

This tension was exacerbated in labour wards, during night duties, or in lower-level facilities where a single health professional was managing multiple responsibilities. Although this challenge was partially associated with mode of birth (e.g. worse in L&D wards), many of the potential causes and solutions are higher up in the health system.

The ratio of staff was higher in OTs, increasing the availability of staff for care provision and RHIS tasks. Participants explained that in the OT, responsibilities for completing individual clinical records and informal OT registers were shared between nurses/midwives, doctors, and other supporting staff.

’In labour [ward], nurses are the ones who fill in the data; in theatre, there is a combination of doctors and nurses. This can cause other data to be missed in our MTUHA.’

Nurse, maternity (District hospital), TZ

This contrasted with task allocation in the L&D ward, where a single staff member was usually responsible for multiple RHIS processes (e.g. completion of registers, medical records, reporting) while simultaneously providing care (often to multiple people, despite the importance of one-to-one care during labour) (Fig 2).

’Sometimes it’s difficult for us to maintain so many registers. Also, the person making the entries finds it really difficult to do it due to the pressure from so many patients and so much work. Sometimes, it’s even hard to complete the reports in time.’

Midwife (District hospital), BD

’After the end of the month, between the 1st and 5th date, we are supposed to draft a report. We have one report. We have a report for the theatre register and another from the labour ward, and we merge them into one report. …the report from labour shows the ones who had surgery or a normal delivery. So, in labour [ward], one book has information from theatre.’

Nurse, OT (District hospital), TZ

The burden was also intensified by monthly reporting cycles because staff were divided between conflicting responsibilities for care provision and data reporting.

’…due to the scarcity of staff, sometimes we forget to enter some information, which becomes a big challenge when closing the monthly report. You might find yourself alone in the ward, or sometimes a mother gives birth quickly, and you hastily write her file and put it aside. Then you go, assist another mother, write her report, and put it aside. When it comes time to enter the data, you might forget to input the information for the first mother, and her data might get lost. A client may deliver at night, but her information can’t be documented until morning.’

Assistant nurse (District hospital), TZ

Staffing shortages impacted data use as well as data collection. Most participants described RHIS data as important and useful. This was embodied by accounts of staying late to complete RHIS duties (n,8) or using personal resources to fulfil tasks (e.g., phone credit). However, most frontline workers conceptualised the importance of data use in terms of their clinical case management, handover, procurement of medical supplies, and accountability (rather than to inform service management, quality of care improvements, changes in population health status etc).

’If you can provide accurate information, even our workload can be visible. For example, if you say you’re busy, but you’ve performed 20 operations, it won’t convince anyone that you have a heavy workload. However, if you say you’re busy and you’ve conducted nearly 200 operations in a month, then it truly appears that there is a significant workload. This can persuade others to consider increasing manpower and providing more medical equipment and medications.’

Doctor (District hospital), TZ

Health professionals expressed the need to prove what they had done, especially with the high volume of cases. In Bangladesh this may be exacerbated by a hierarchical medical model, as explained by a midwife:

’I have to keep documents of what I am doing. If I said through my mouth that I did ten deliveries in the last month, would you believe me? Would the government believe me? I do my job, and there is a document showing that. When you see that, you will understand.’

Midwife, labour ward (Upazilla), BD

The testimonies highlight a conflict between needing data for accountability, to create evidence for staffing increases, and for protection (e.g. MPDSR, complaints).

Organisational factors were determined by facility or higher-level policies, resourcing, and culture (rather than mode of birth)

Factors such as governance, training, and supervision were similar for vaginal and caesarean births in both countries. This was partly because data for caesarean births were collected and aggregated in the same L&D ward registers as vaginal births. For example, many participants prioritised the completeness and timeliness of register filling over other dimensions of RHIS data quality and use.

’Sometimes it’s just the attitude of people because they fill in just to avoid being told they haven’t filled in. So, they fill in just to appear as if they’ve done the job…. Sometimes, they are not filled in accurately, but they are filled in just on time’.

Doctor (District hospital), TZ

’Our supervisor, sir, handles this with his own hands, and he tells us how to do it in a proper way…. They tell us to fill up all the cells in the register and they check in to this. They also cross-check the total summary and the report.’

Midwife (Upazilla), BD

This was reinforced by data reporting structures and supervision approaches. Supervision was top-down with a strong focus on quality monitoring, especially the completeness of registers and tally sheets (rather than localised data use), which negatively impacted on participants sense of data ownership and wellbeing. Participants wanted more positive encouragement.

’They come and then look at all the registers. They check if we are doing our thing right… They used to come a lot. They used to catch our mistakes often and made us listen to their harsh words because of the mistakes we made.’

Family Welfare Visitor (assistant nurse, Maternal and Child Welfare Centre), BD

Data use varied by role, level of responsibility and training, but frontline health professionals were also overstretched with practical limitations on their capacity to take on additional tasks. Programme and facility managers were more likely to discuss data use for public health or higher-level service planning. In Bangladesh, the strong hierarchical structures meant that lower-level and frontline staff were not empowered for local data use. This was generally perceived as being the responsibility of more senior colleagues.

’Our job is to submit the report. We don’t know what further is done with it. We provide a monthly report to the statistician, and after the data entry, he sends it where needed. We don’t have any other authority over these data.’

Midwife, labour ward (District hospital), BD

This was reinforced by participant accounts around in-service training. Increasing access to training was valued by many, but health professionals generally expressed interest in training on data collection and register filling (rather than interpretation and use). These organisational factors were determined by facility or higher-level policies, resourcing, and culture (rather than mode of birth).

Strengths and limitations

The participation of 79 health professionals, managers, and policymakers from public health services (and 22 different facilities) in Bangladesh and Tanzania ensured a diversity of experiences and perspectives were represented in the data. The analysis revealed that the effect of mode of birth on RHIS data processes was most evident at the point of data collection (within facilities). The themes were also powerfully coherent between Bangladesh and Tanzania, despite differences in national context and health system structure. Our qualitative research was undertaken as part of a study to develop and test the EN-MINI tools. Quantitative data (using the EN-MINI tools) were therefore collected from the same facilities and offices at the same time; they showed strong performance in data quality and use at district-level, with opportunities for focused action in health facilities [32].

As a qualitative study, we intended to contribute to a deeper understanding of how and why mode of birth effects routine data within the context of Bangladesh and Tanzania, rather than produce broadly generalisable findings. Bangladesh and Tanzania enjoy high government engagement in improving RHIS data, and both countries participated in the previous EN-BIRTH studies regarding MNH measurement [14, 32]. Furthermore, our study focused on public health settings, but many births around the world happen at home or in private facilities, including in Bangladesh where 45% of births are in private facilities [64]. Non-public service providers need to be included in both the research and policy making, especially for issues such as caesarean section, which is much higher in the private sector (Table 2) [39, 40].

Finally, we did not include service-users in our study; however, their experience of care and perceptions of data are important and under-represented in the literature. We need to better understand the potential impact of RHIS data processes on the woman and newborn’s experiences and connections with care providers; ultimately, the data belongs to them, and we must ensure that RHIS processes meet their needs and expectations (Table 2).