Designed to Fail? Revisiting Uganda’s Maternal Health Policies to Understand Policy Design Issues Underpinning Missed Targets for Reduction of Maternal Mortality Ratio (MMR): 2000-2015

Background: Despite Uganda and other sub-Saharan African countries missing their maternal mortality ratio (MMR) targets for Millennium Development Goal (MDG) 5, limited attention has been paid to policy design in the literature examining the persistence of preventable maternal mortality. This study examined the specific policy interventions designed to reduce maternal deaths in Uganda and identified particular policy design issues that underpinned MDG 5 performance. We suggest a novel prescriptive and analytical (re)conceptualization of policy in terms of its fidelity to ‘3Cs’ (coherence of design, comprehensiveness of coverage and consistency in application) that could have implications for future healthcare programming. Methods: We conducted a retrospective study. Sixteen Ugandan maternal health policy documents and 21 national programme performance reports were examined, and six key informant interviews conducted with national stakeholders managing maternal health programmes during the reference period 2000-2015. We applied the analytical framework of the ‘three delay model’ combined with a broader literature on ‘policy mixing.’ Results: Despite introducing fourteen separate policy instruments over 15 years with the goal of reducing maternal mortality, by the end of the MDG period in 2015, only 87.5% of the interventions for the three delays were covered with a notable lack of coherence and consistency evident among the instruments. The three delays persisted at the frontline with 70% of deaths by 2014 attributed to failures in referral policies while 67% of maternal deaths were due to inadequacies in healthcare facilities and trained personnel in the same period. By 2015, 37.3% of deaths were due to transportation issues. Conclusion: The piecemeal introduction of additional policy instruments frequently distorted existing synergies among policies resulting in persistence of the three delays and missed MDG 5 target. Future policy reforms should address the ‘three delays’ but also ensure fidelity of policy design to coherence, comprehensiveness and consistency. This paper combines Thaddeus and Maine’s ‘three delays’ model with policy mix design approach.19,26 In the 3D model, Thaddeus and Maine convincingly argued that the prevention of maternal mortality is largely dependent on the length of the interval between the onset of obstetric complication and its outcome. Prompt and adequate treatment promotes positive outcomes while delayed treatment adversely affects outcomes. Delay I refers to delays in seeking care.27,28 Delay II involves delays in accessing a health facility with functional EmOC services; and Delay III involves the delay in accessing appropriate EmOC services once at the facility, or due to subsequent referral from one health facility to another (ie, delays related to correct diagnosis of complications and appropriate action and, delays in the referral pathways).9,29-31 According to 3D, the delays in seeking care, reaching a health facility and receiving appropriate care feed into each other. For example, the delay to seek care is affected by how far a pregnant woman has to travel for healthcare (Delay II) but also the mother’s perception of the quality of services in respective health facilities (Delay III). In addition, the three delays can occur in a vicious cycle with delay I leading to delay II, which in turn leads to delay III as the cycle continues. The paper derives from the broader social science literature on policy design as understood in two principle ways: First, procedurally as a process of selecting the most appropriate policy instruments to solve a problem or achieve a policy goal,19 Second, as a substantive output of the policy process – the content, instruments and goals achieved.32 Policy instruments refer to “the means by which government policies are carried out.”33 Policies may well arise from an objective process of problem analysis and rational selection of appropriate instruments. However, they could also arise out of a negotiation or clash of interests.34 We position ourselves here as ‘rationally substantive’ where policy design is considered as an outcome of an objective process that combines policy instruments to achieve the goals sought. An effective policy design basically is one that is able to solve the problem to which the policy is responding.19 Borrowed from economics, the term ‘policy mix’ draws attention to how blending different policy instruments can affect intended policy outcomes in unexpected ways. The literature tends to depict failure to achieve outcomes as either due to inadequate policy design; or resulting from poor policy instruments’ interaction.26 We assessed Uganda’s adopted policy instruments in terms of their temporal stability (ie, over a fifteen-year period) and in terms of the extent to which they conform to what we introduce as the 3Cs: coherence – the synergies or logical linkage of policy instruments to the stated goal; their comprehensiveness – the completeness of instruments in regard to the three delaysand consistency – the absence of contradictions or the extent to which policy instruments work together towards the same goal.19,31,35 These relationships are shown in diagrammatic form below (see Figure 1). Conceptual Framework. A retrospective mixed methods study was conducted at national level covering policies regarding maternal health. Uganda was treated as a single unit. Retrospective policy analysis was utilised as a powerful tool to inform future reform via reflection on past performance.36 Data were collected in two phases between April and July 2018. The first round involved document review triangulated by interviews with national policy-makers in the second round. Documents reviewed were retrieved from the MoH Knowledge management portal (http://library.health.go.ug/publications) and the libraries of MoH, Makerere University and the Ugandan Parliament. Two main types of documents were assessed: (a) Maternal Health Policy Documents. A general search was conducted using the following strategy: “maternal health” AND “Reproductive Health” AND Policy OR Guidelines OR Standards OR Strategy OR Plan. The inclusion criteria were: (i) A document, which explicitly spelt out the objective of reducing maternal mortality; (ii) Covering the period 2000-2015, (iii) Authored by the government of Uganda. We excluded documents that did not directly state a policy on maternal health mortality and mapped 14 policies, goals, targets and interventions from 16 policy documents. Using a document review checklist, we extracted raw text on policy provisions, duration of the policy, goal, objectives/outcomes, targets and interventions/policy instruments for each of the three delays. (b) Government progress reports covering performance on various maternal health indicators in the MDG period. These included; Annual health sector performance reports (n = 11), Maternal and Perinatal Death Reviews (n = 3), UDHS (n = 4), Health Centre IV (HC IV) and Hospital Census (n = 1), and National Service Delivery Survey (n = 2). We extracted trend data for 15 years on the following indicators that reflect the three delays: access to maternal health services ie, population within 5 km to the nearest health facility (delay two), functionality of EmOC, stock levels of essential Reproductive Health (RH) medicines (delay three) and reference to any of the three delays in recorded maternal deaths (this could reflect any of the three delays). We then extracted corresponding narratives that helped interpret the statistics. Key informant interviews (n = 6) were conducted with selected national respondents. These were supplemental and served to clarify our observations from document review. All participants had more than 10 years’ service and were involved in national maternal health policy processes between 2000 and 2015. They included: a researcher (n = 1), national maternal health programme managers (in service or retired) in the MoH (n = 2), an Obstetrician/Gynaecologist (n = 1) and consultants hired to develop or evaluate maternal health policies (n = 2). These brought vivid experiences from their direct involvement in policy design, implementation and evaluation to illuminate the observed performance with respect to the three delays. An interview guide was configured to elicit their perspectives on the different policy packages and what they thought influenced their performance. We obtained written informed consent from participants. All interviews were audio-recorded, transcribed verbatim and reviewed for consistency. Confidentiality was maintained by anonymising the transcripts using generic identifiers such as, “researcher, gynaecologist/ obstetrician and national programme manager.” All data was stored electronically under secret password protection only known by the corresponding author. The application of the 3Cs framework in assessing effectiveness of design is underpinned by the assumption that; (1) there is a clearly defined problem (in this case high maternal death due to the three delays), (2) the interventions/ policy instruments/solutions to alleviate the problem are known (in our case, the prescriptive interventions for the three delays outlined by Calvello et al,37 and, (3) there is a theory of change linking solutions to the problem – the three delay model.11,38 Then fidelity of design to the 3Cs was assessed by asking the questions; 1) do the policy instruments speak to the problem (Coherence)?, are all the policy instruments needed to alleviate the problem included (Comprehensiveness)? and, are the policy instruments mutually reinforcing/do they speak to each other (Consistency). We undertook a deductive manifest content analysis to assess policy design following three steps: Step 1) the extraction of policy instruments and interventions. Step 2) for each policy, interventions were grouped modelling the 3D framework. Step 3) each intervention/ instrument was tabulated against each of the 3Ds and mapped ( Tables S1 and S2, Supplementary file 1). Using our 3Cs framework for assessing policy mixes, a two-step analysis process was performed to assess policy design. Step 1: comprehensiveness was quantitatively assessed by assigning a score of 1 or 0 for the presence or absence of an instrument for each delay in comparison with prescriptive interventions for the three delays.37 The expected score was 8 to be able to conclude the policies put in place all the instruments to address the three delays. The total score was the actual score arising from the scoring of presence or absence of expected interventions. From the scores, proportions were computed by comparing aggregate scores per policy with the expected score of 8 across the MDG period (Figure 2) to observe a trend in comprehensiveness over the 15 years. Step 2: Coherence was assessed guided by the question; “do the instruments speak to the problem?” Consistency was assessed by answering the question; “are there observable contradictions or linkages among the policy instruments?” Interventions for each of the three delays as operationalised in a publication by Calvello et al37 were tabulated against interventions proposed in policy documents over the period 2000-2015. The corresponding author assessed through the surface (manifest) observation39 the presence and lack of coherence and consistency among policy instruments within and across the three delays over the 15 years being assessed. The presence or lack of coherence and consistency were colour coded green and red respectively (Supplementary file 2). The results were independently assessed using the same approach by three senior faculty close to the study and two senior faculty distal to the study based out of the Ugandan context. The corresponding author revised the table based on the feedback from the senior faculty and shared the final results which they approved of. Since interviews were few and focused on specific observations from the documentary review, we directly corroborated results from this with explanatory narratives from interviews as manually extracted and applied to enhance the interpretation of results conducted via our document review. To assess the manifestation of 3Ds we extracted data related to distance (population within 5 kilometres to the nearest health facility), functionality of EmOC, stock levels of essential RH and avoidable factors for maternal deaths. Illustrative quotes were extracted from the interviews to interpret and explain the results. Testing the Comprehensiveness of 3D Policy Design. Abbreviations: CEmOC, comprehensive emergency obstetric care; MSRH, maternal sexual reproductive health; EmOC, emergency obstetric care; BEmOC, basic emergency obstetric care; TBAs, traditional birth attendants; HC IV, Health Centre level four. * Ongoing policies by the time of this study.