Aligning vertical programmes with health systems: a case study of the HIV programme at district level in South Africa

Abstract

Rationale It is widely recognised that population health can be improved by strengthening health system capacity to deliver health services that tackle a wide range of diseases and that people can use when they need them. However, many low- and middle-income countries (LMICs) have established disease control programmes (DCPs) that employ dedicated machinery (e.g. health workforce, drug delivery, health infrastructure) to deliver health services that tackle a specific disease (traditionally termed the ‘vertical’ approach). DCPs have beneficial (increased coverage of disease control interventions) as well as detrimental effects (hamper holistic care because patients are seen on the basis of their disease, increase duplication and fragmentation in service delivery, and draw scarce resources from overall health systems that are designed to tackle a wide range of diseases). Integrating DCP services within overall health systems (traditionally termed the ‘horizontal’ approach) is thus increasingly proposed. Some suggest that DCPs should also be integrated within other health system functions (e.g. planning, financing, monitoring and evaluation [M&E]) in order to minimise fragmentation in managing health services. With increasing interest in integration, research is needed to inform discussions on how to achieve it. Since integration is not necessarily about abolishing all elements of ‘verticality’ but rather about finding ways in which beneficial aspects of DCPs can exist with health systems in a fruitful symbiosis, research is needed on how different extents of DCP integration may affect health systems. Understanding the extent of integration is important as the acceptable degree of ‘verticality’ may vary across different contexts. In South Africa several DCPs exist – notably HIV, tuberculosis (TB), maternal and child health (MCH) – but it is official government policy to integrate them within the district health system. As such, DCP services (e.g. HIV testing, TB diagnosis and treatment) are supposed v to be integrated at the point of care – provided by multi-skilled health workers through multi-functional health facilities. Policy also advocates integration at managerial level – in a context where district managers have been delegated the authority for implementing health services (planning, supervision, monitoring). As such, DCP managers should relinquish responsibility for DCP interventions to district managers to whom they should provide specialist support. There is however no policy guidance on how to implement integration. Existing policy also makes no mention of DCP integration within district health system functions such as planning or M&E, and yet it seems important that district managers have administrative authority over these in order to effectively manage integrated services at district level. Evidence to inform implementation guidance is limited. While there is some research on service integration at the point of care, integration at managerial level (administrative integration) is less understood. The limited available evidence suggests there has been little progress with achieving administrative integration, but there is no research exploring why. For example, administrative integration may require changes to health system organisational structure (how roles are allocated and lines of authority), and culture (actor attitudes, behaviours and values). However, whether the prevailing health system organisational structure and culture would support integration has not been researched. Further, collaboration (communication, joint working) between programme and district managers is necessary for integration to succeed, but the extent to which it happens is little researched. This PhD addresses some of the afore-mentioned gaps in understanding by measuring the extent of integration between DCPs and the district health system in South Africa, and exploring how organisational structure and culture may influence integration. The four studies that make up this PhD explore these questions using the HIV programme as an vi example of a DCP and monitoring and evaluation (M&E) as an example of a health system function. Aims The aims of this PhD are: a) to explore the use of methods for measuring the nature and extent of HIV programme integration within district health system M&E in South Africa (studies 1 to 3); and b) to explore the influence of the health system organisational structure and culture on integration (study 4). HIV M&E is a system (people, technology, processes, and management structures) for producing HIV information and using it for monitoring HIV services. In the absence of policy guidance defining what M&E integration means, this research hypothesises a model of HIV M&E integration that has the following characteristics: a. Operational integration: processes, technology and personnel for producing HIV information (collecting, collating, analysing and disseminating) are integrated within the district health information system (DHIS); b. Administrative integration: district managers exercise administrative authority over HIV M&E – oversee HIV data collection and collation (check data quality, compile and submit reports) and use HIV data for monitoring progress with implementation of HIV interventions in districts; and DCP managers at sub-national level provide specialist support (e.g. on technical aspects of data quality, and interpretation of HIV data). c. Collaborative actor relations: district and sub-national DCP managers share task-related communication (talking one-on-one about HIV M&E tasks), and attend the same committees where HIV data are discussed and used for monitoring HIV services. vii Methods The research was conducted in two of nine provinces (one urban and one rural) during 2009 to 2012. One district per province was studied. Quantitative and qualitative data were collected in three phases through: a) interviews with 51 health managers located at sub-national level (health facility, sub-district, district and provincial) which included: district, programme (HIV, TB and maternal and child health) and health information managers; b) interviews with eight participants at national level (HIV programme, health information and health system managers); c) document reviews (policies, health plans, operating procedures, M&E tools and documents); and d) audits at 11 health facilities. Given the dearth of methods for measuring the extent of integration, the research adapts and applies existing health system research methods in new ways and uses methods traditionally used in organisation and social science research and applies these methods to health systems research. To assess operational integration, an existing analytical framework – developed by Atun and colleagues for measuring the extent of integration of DCPs within health system functions – was adapted and applied to rate the extent to which HIV data collection and collation forms and processes for reporting, analysing and disseminating HIV data were integrated within the DHIS as ‘no integration’, partial’ or ‘full integration’. To assess administrative integration, Bossert’s decision-space analysis – traditionally used to assess if and how managers exercise authority over health system functions following decentralisation – was adapted to quantify the degree of exercised authority over HIV M&E (the extent to which managers perform HIV M&E tasks) as ‘low’, ‘medium’, or ‘high’. The degree of exercised authority and HIV M&E knowledge were compared between district and programme managers. The extent of communication was quantified using social network analysis (SNA). SNA measures were computed to describe: actor centrality (identify actors with the lowest and highest number of communication links to others, and those who connect viii otherwise disconnected actors); density (quantify cohesiveness of task-related communication); and homophily (quantify extent of communication within versus outside manager groups). Block modelling was applied to identify management committees that link programme and district managers. Finally, Mintzberg’s organisational configurations framework was applied to describe three organisational parameters of the health system: a) the type of decentralisation (whether the locus of decision-making about the design of the HIV M&E system lies at higher or district level); b) the key part of the organisation (whether sub-national programme or district managers are the key role players in HIV monitoring); and c) coordination mechanisms used (highly formalised versus output-based mechanisms). The study then analysed how the observed organisational configuration influence integration. Key findings The results show that operationally the HIV M&E system has two separate sub-systems. One produces information only on anti-retroviral treatment (ART) services, is not integrated within the DHIS and limits availability of ART data at district level. The second produces information on non-ART services (e.g. HIV counselling and testing or prevention of mother to child transmission) and some aspects of it are not integrated (data collection forms) while others are partially (personnel) or fully integrated (software). District managers exercise high degrees of administrative authority over HIV data collection and collection, but there is duplication as programme managers perform some of these tasks as well and seldom perform HIV M&E specialist support roles (partial integration). HIV data use is not integrated as: district managers (many of whom have low HIV M&E knowledge) exercise low degrees of authority in using HIV data; while programme managers (usually with high HIV M&E knowledge) exercise high degrees of authority and use HIV data in silos excluding district managers. ix In both sites task-related communication networks of all managers are moderately cohesive, but provincial HIV programme managers as a group seldom talk to the district managers to whom they should provide specialist support. Though several management committees discuss and use HIV data for monitoring, few connect district and programme managers to potentially foster joint monitoring. Finally, the health system organisation is characterised as Mintzberg’s machine bureaucracy that is incongruous with integration policy objectives. It is centralised (district actors play a peripheral role in decisions regarding HIV M&E design); highly formalised (use rules to enforce compliance and control how M&E work gets done rather than defining outputs that should be achieved); . The organisational culture promotes programme managers as the lead role players (district actors not valued as key players, investment in capacitating programme and not district managers), and leadership styles fail to foster collaborative relations amongst programme and district managers. Implications High degrees of ‘verticality’ were observed which potentially hamper prospects of integrated health services within the district health system. First, the ART M&E system limits availability of HIV data at district level and district managers oversee the production of HIV information while programme managers use it, which limits the extent to which district managers can manage health services in a holistic manner. Setting up a parallel system to fast-track data submission to higher levels is a missed opportunity to strengthen DHIS capacity to produce HIV (and other) data in a timely manner. Second, provincial programme managers largely communicate amongst themselves and seldom with the district managers for whom they should provide specialist M&E support. This means the HIV M&E expertise resides at provincial level and is not available at district level where it is needed. Third, a centralised and highly formalised health system that promotes and values programme x managers as key role players in programme operations at district level undermines policy intentions of district managers assuming this leadership role. This research identifies where interventions can be targeted to achieve higher degrees of integration for district health system and potentially patients’ benefit, and recommends organisational change that is needed to better enable translation of stated policy intents into practice. Lessons learned from this specific HIV M&E case may be relevant for the integration of other programmes in South Africa, and applicable to other LMICs, particularly those seeking to integrate programmes within the health system at district level.