On people, data and systems : perspectives on routine health data processing and its digitalization in Tanzania

Abstract

Background: Facility-based routine health information is captured in health management information systems by health care providers and is the main data source for health system planning and outcome monitoring in Tanzania and other low- and middle-income countries. While this system is fully digitalized in high-income countries, it is still partly paper-based in others. These use i) facility registers, ii) daily tally sheets and iii) monthly summary forms, which are later entered into the District Health Information System-2 software. These hybrid systems are prone to errors related to i) data entry, ii) calculation and iii) data transfer, with negative implications for data completeness and availability. The unavailability of data and lack of trust in its quality may lead to low data use for resource forecasting and planning, especially at subnational and facility levels. Through automatization of data processing, digital technology may be able to address these challenges, making it especially attractive in settings with high disease burdens and few resources. One example of a promising digital solution for low-resource settings is Smart Paper Technology, which produces automated electronic registers and summary reports by scanning bar-coded forms from individual service encounters. Implementation research, however, suggests a complex interplay between the implementation environment and the introduction and sustained institutionalization of technology.

The aim of this thesis was to understand the social practices involved in generating and processing routine maternal and newborn health data, using paper-based and digital tools within the health management information system in Tanzania. Smart Paper Technology and the current health management information system with its different digital components are used for evaluation.

Study I had the objective of understanding health care providers’ and facility/district managers’ perceptions of Smart Paper Technology and to assess time spent on documentation with the new system. A time-motion study, before and after the introduction of the technology, was applied together with eight focus group discussions with 18 health care providers from three health facilities and 11 in-depth interviews with healthcare managers from one district authority. Quantitative data was analyzed using descriptive statistics and bivariable modelling. Reflexive thematic analysis was used to analyze qualitative data. Findings illustrate challenges to Smart Paper Technology implementation related to pre-existing health system bottlenecks, e.g. lack of human resources, supervision and transport, but also a difference in values assigned to the new system by health care providers and their managers. Health care providers found Smart Paper Technology useful and applicable to their context with perceived benefits for documentation and clinical care. These experiences were confirmed by quantitative data, showing no significant difference between time spent on overall documentation pre- and post-introduction of Smart Paper Technology (27 vs 26 %, adjusted p 0.763) but an increase in time spent on clinical tasks (26.9 vs 37.1%, adjusted p 0.001). Health care managers, in contrast, found it difficult to identify benefits from the new technology for their own work related to national reporting, due to access problems with the digital dashboard and questionable quality of Smart Paper Technology data. They therefore continued to focus managerial efforts on the existing health management information system.

Study II’s objective was to assess the quality of Smart Paper Technology data for maternal care services related to i) completeness and timeliness and ii) internal consistency. A cross-sectional survey over 12 months was performed in 13 health facilities using data from the Smart Paper Technology system and District Health Information System-2. Descriptive statistics were produced based on indicators derived from the World Health Organization’s Data Quality Review Toolkit. Results show that data quality of the Smart Paper Technology system was not superior to that of the pre-existing health management information system overall. This may be linked to the effects of duplicate data entry on health care provider performance and consequently on data completeness. Smart Paper Technology performed slightly better in some aspects of internal consistency: Fewer health facilities produced only one or two outliers with Smart Paper Technology in each month of the study period (antenatal care=4, care during labour = 6, postnatal care =4) than with the District Health Information System-2 (antenatal care= 7, care during labour= 9, postnatal care= 6). Smart Paper Technology also yielded higher consistency for the documented postpartum use of oxytocin in relation to the number of documented deliveries with 62% of facilities showing a less than 10% difference between these indicators as opposed to 38% for the District Health Information System-2. However, the pre-existing system demonstrated better data quality in all other quality dimensions, i.e. data completeness, timeliness and consistency of data trends over the study period.

Study III: The objective was to improve understanding about the processes involved in health care providers’ data use; which type of information is used together with health management information system data and for what purposes. A constructivist grounded theory-based ethnographic approach was applied, consisting of i) 14 in-depth interviews with health care providers from maternity wards in two hospitals, as well as ii) 48 hours of observation in the maternity wards and ii) two focus group discussions with 11 health care providers from the same hospitals. Findings illustrate how health care providers appropriated numeric data from the official health management information system and narrative data that they had produced for clinical documentation to safeguard social relationships with superiors, patients and the community they served. While they identified themselves as data collectors and not users of the health management information system, they applied narrative clinical documentation systems to service improvement and to protect themselves against litigation or managerial reprimands.

Study IV’s objective was to generate knowledge on experiences and perceptions of health care policymakers in Tanzania related to data, data systems and the implementation of digital technology to support health information management. 16 in-depth interviews with healthcare managers from national and subnational levels were conducted and analyzed using reflexive thematic analysis. Results suggest that the health management information system in Tanzania is governed using institutional and discretionary power. Institutional power was mainly used at the national level to conceptualize data collection and processing systems and the scale-up of digitalization. Discretionary power was mainly used for implementation at subnational level. The use of different power practices was influenced by available funding and health care managers’ perception that health care providers, the primary data collectors, lack motivation to perform and are unpredictable in their actions regarding the continuous production of good data quality.

Conclusions: Acceptance or rejection of digital technology was influenced to a considerable extent by social practices at all levels of the health system. These included actors’ perceived benefits of maintaining existing social practices. These practices, which are part of an organization’s culture related to data and data processes, require attention during the conceptualization and implementation of health information systems. Numeric and contextual information is used concomitantly at various levels of Tanzania’s health management information system. The health management information system in Tanzania forms a complex adaptive system with inherently high levels of unpredictability, non-linearity, self-organization and adaptation over time. Health care managers’ power practices in the conceptualization and implementation of policies reflect this complexity. Contextual factors affect digital technology integration and have consequences for data quality and use of digital AND paper-based health management information systems. Context may therefore be even more important than the format and technology of data collection and processing.