Providing oxygen to children in hospitals: a realist review

Abstract Objective To identify and describe interventions to improve oxygen therapy in hospitals in low-resource settings, and to determine the factors that contribute to success and failure in different contexts. Methods Using realist review methods, we scanned the literature and contacted experts in the field to identify possible mechanistic theories of how interventions to improve oxygen therapy systems might work. Then we systematically searched online databases for evaluations of improved oxygen systems in hospitals in low- or middle-income countries. We extracted data on the effectiveness, processes and underlying theory of selected projects, and used these data to test the candidate theories and identify the features of successful projects. Findings We included 20 improved oxygen therapy projects (45 papers) from 15 countries. These used various approaches to improving oxygen therapy, and reported clinical, quality of care and technical outcomes. Four effectiveness studies demonstrated positive clinical outcomes for childhood pneumonia, with large variation between programmes and hospitals. We identified factors that help or hinder success, and proposed a practical framework depicting the key requirements for hospitals to effectively provide oxygen therapy to children. To improve clinical outcomes, oxygen improvement programmes must achieve good access to oxygen and good use of oxygen, which should be facilitated by a broad quality improvement capacity, by a strong managerial and policy support and multidisciplinary teamwork. Conclusion Our findings can inform practitioners and policy-makers about how to improve oxygen therapy in low-resource settings, and may be relevant for other interventions involving the introduction of health technologies.


Introduction
Oxygen is an essential medical therapy that has been saving lives for over 100 years. 1 Oxygen therapy is used not only for pneumonia and other primary lung diseases but also many other conditions that result in hypoxaemia, such as sepsis, severe malaria, status epilepticus, trauma, obstetric and neonatal conditions (respiratory distress, apnoea, asphyxia, sepsis), surgical care and anaesthesia. A systematic review estimated that, globally, hypoxaemia affects about 13% of children admitted to hospital with pneumonia, about 20% of sick neonates and 10-15% of children admitted with conditions such as malaria, meningitis or convulsions. 2 Given that hypoxaemia is a major risk factor for death, 2,3 oxygen therapy is important for improving child health outcomes.
Effective oxygen therapy requires prompt and accurate detection of hypoxaemia and appropriate administration of oxygen, combined with good clinical evaluation and management of the underlying condition. 4 Improvements in the technology and affordability of pulse oximetry -the standard method for detecting hypoxaemia -are enhancing its accessibility for hospitals in low-resource settings. 1 Oxygen may be supplied by oxygen cylinders, oxygen concentrators or larger oxygen plants, each of which have unique advantages and disadvantages, particularly when used in hot, humid or dusty environments. 1,5 The World Health Organization (WHO) has produced guidelines on the clinical use of oxygen [6][7][8] and oxygen equipment. 9 Despite these advances, the availability of pulse oximetry and oxygen supplies remains limited in regions of the world where they are most needed. 10 Furthermore, workforce limitations and health-system failures limit the ability to maintain, sustain or effectively use oxygen even when it is available. 1 A solution to this must be multifaceted, and is likely to be context-specific.
We aimed to identify and describe interventions to improve oxygen therapy in hospitals in low-resource settings, and to determine the factors that contribute to the success or failure of interventions in different contexts.

Methods
We used a realist review approach 11 to study not only whether oxygen therapy interventions work, but also how and why complex programmes work in particular contexts and settings. Realist review is a theory-driven systematic review method that involves identifying key mechanistic theories about how projects might work, searching the evidence about project implementation and impact (including variability between contexts), and then testing the evidence with respect to the theories. [11][12][13][14] Our review was prospectively registered on the PROSPERO register of systematic reviews (CRD42015032405).

Identification of theories
We made a preliminary scan of the literature to identify potential theories to explain how improved oxygen therapy systems could impact on clinical outcomes. This resulted in a list of candidate theories, each describing a mechanism through which the intervention influences particular outcomes in particular contexts. We consulted key experts and stake-Objective To identify and describe interventions to improve oxygen therapy in hospitals in low-resource settings, and to determine the factors that contribute to success and failure in different contexts. Methods Using realist review methods, we scanned the literature and contacted experts in the field to identify possible mechanistic theories of how interventions to improve oxygen therapy systems might work. Then we systematically searched online databases for evaluations of improved oxygen systems in hospitals in low-or middle-income countries. We extracted data on the effectiveness, processes and underlying theory of selected projects, and used these data to test the candidate theories and identify the features of successful projects. Findings We included 20 improved oxygen therapy projects (45 papers) from 15 countries. These used various approaches to improving oxygen therapy, and reported clinical, quality of care and technical outcomes. Four effectiveness studies demonstrated positive clinical outcomes for childhood pneumonia, with large variation between programmes and hospitals. We identified factors that help or hinder success, and proposed a practical framework depicting the key requirements for hospitals to effectively provide oxygen therapy to children. To improve clinical outcomes, oxygen improvement programmes must achieve good access to oxygen and good use of oxygen, which should be facilitated by a broad quality improvement capacity, by a strong managerial and policy support and multidisciplinary teamwork. Conclusion Our findings can inform practitioners and policy-makers about how to improve oxygen therapy in low-resource settings, and may be relevant for other interventions involving the introduction of health technologies. Systematic reviews Improving oxygen therapy for children Hamish Graham et al. holders, including interviewing the authors of five large-scale oxygen therapy projects. Interviews were recorded and transcribed for accuracy, but were not formally analysed and were used only to assist in identifying emerging themes.

Search strategy
We made a systematic search of online databases (MEDLINE®, Embase®, CI-NAHL, AIM, LILACS, the Index Medicus for the Eastern Mediterranean Region, the Index Medicus for South-East Asia Region, the Western Pacific Region Index Medicus, CAB Global Health, Health Systems Evidence, PubMed® (for e-publications) and Google Scholar (first 500 citations)) on 10 August 2016. We searched variations of keywords "child", "oxygen concentrator", "oxygen cylinder", "oxygen therapy", "oxygen delivery", "oxygen administration" and "developing country" (MEDLINE® search; Box 1). We also searched websites (e.g. WHO and the International Union Against Tuberculosis and Lung Disease), contacted oxygen therapy experts and reviewed the reference lists of included studies to identify additional published and unpublished studies.
We included any evaluation of an improved oxygen therapy system involving a hospital in a low-or middleincome country (World Bank definitions at the time of the study). An improved oxygen therapy system included introduction of an improved oxygen source (oxygen concentrator, cylinder or plant) with or without: other equipment (e.g. pulse oximeters, oxygen delivery devices); education (e.g. training materials or visits); health-system or quality improvement activities (e.g. financing, supply chain, supervision). Two investigators independently reviewed the titles, abstract and full-text of the studies for inclusion, with the adjudication of a third reviewer where consensus could not be reached.
We assessed the quality of each study on two criteria: relevance (i.e. addresses the candidate theories about how improving oxygen systems lead to improved clinical outcomes); and rigour (i.e. provides credible data to reach a conclusion). Most studies contributed to the testing of a particular theory more than others. We only reported results where there were credible data.
To assess the quality of studies reporting clinical outcomes we used the Effective Public Health Practice Project quality assessment tool for quantitative studies. 15,16

Data extraction and synthesis
We used a data collection form adapted from the Cochrane Effective Practice and Organisation of Care (EPOC) group 17 and process evaluation 18 tools. For each project we extracted data on the theory (implicit or explicit); context (e.g. facility characteristics, oxygen capacity and needs); interventions (using Cochrane EPOC categories); processes (e.g. quality, fidelity to the project plan, how well it reached the intended beneficiaries); and outcomes (clinical, quality of care, cost, equipment function).
The data were recorded in a series of tables, enabling possible mechanistic theories to be explored within individual projects and between projects. After repeatedly exploring the partially developed theories, we aggregated them into major theoretical themes. We then analysed the available data to support, negate or refine the theories, thereby identifying key factors affecting the success of projects.

Results
Of 2433 records screened, 72 full-text articles were assessed for eligibility (Fig. 1 (15) used non-comparative evaluation methods, one project used a contemporaneous control and four used a historical control.
The earliest projects (in the 1980s) introduced oxygen concentrators primarily for anaesthetic purposes, and evaluated their technical function and cost implications (Table 1). Since the late 1980s, seven large-scale paediatric oxygen projects have been evaluated, involving complex interventions targeting technicians, clinicians and often administrators and policy-makers (Table 2). Recently, four projects have explored the utility of improved power supplies for oxygen concentrators, including solar power systems (Table 1  and Table 2). Four large-scale effectiveness studies demonstrated reduced in-hospital mortality from pneumonia and other hypoxaemic conditions among children aged < 5 years follow-  ing an intervention to improve oxygen therapy (Table 3). Due to heterogeneity in reported outcomes and insufficient comparable data, a meta-analysis was not possible. Overall, projects used a variety of strategies to improve oxygen therapy. We identified five themes (Box 2), each containing multiple mechanisms that explain how interventions to improve oxygen therapy work (or not) in different contexts. We tested these theories using data from all included projects and identified key factors reducing or enhancing the efficacy of oxygen therapy projects (Box 3).

Oxygen access
Increasing the availability of oxygen therapy to patients was an implicit aim in virtually all the intervention projects we studied, but few studies evaluated it. Hospitals in most projects had some access to oxygen cylinders (Gray et al.), 21,22,[24][25][26]35 or concentrators 50 before the intervention, but supply was limited by depletion of cylinders (which are costly and difficult to refill) and broken equipment. The only baseline data on oxygen availability to patients were from Papua New Guinea, where oxygen was available for 326 (87%) of 375 children who had hypoxaemia on admission. 56 Two large-scale projects reported increased access to oxygen cylinders or concentrators compared with the baseline. 46 A prospective evaluation reported on equipment and clinical outcomes. 31  (. . .continued) oximetry) at baseline achieved greater reductions in pneumonia case-fatality rates post-intervention than hospitals with reasonable access at baseline. 52,56 We identified three major determinants of oxygen availability: (i) equipment; (ii) maintenance; and (iii) affordability.

Equipment
Although the provision of quality, userfriendly equipment is necessary for an improved oxygen therapy system, it is not easy to achieve. Baseline data from several projects documented poor quality and broken equipment that had often been donated by international donors but did not suit the needs of its users or the environmental conditions. 29,[48][49][50]53,59 Studies consistently reported the importance of procuring oxygen concentrators that were proven to work in hot and humid conditions. This was based on observations of maintenance problems and premature equipment failure when projects provided untested concentrators or used a variety of concentrator types (Gray et al.). 28,36,46,51,53 Pulse oximeters were not used in most of the oxygen projects before 2005.
Where pulse oximeters were used, they were valued as both a diagnostic aid (determining which children require oxygen therapy) and therapeutic aid (giving staff and patients confidence in oxygen therapy; Gray et al.). 3,49,52,56 Limited data suggested that the introduction of pulse oximetry can improve oxygen use and reduce pneumonia case fatality rates, even with suboptimal oxygen access. 3,56 Oxygen delivery devices varied between projects. Nasal prongs were easiest for health-care workers, and some projects reported confusion when multiple options were provided to inexperienced users. 24,35,37,46 Some projects used flow-splitters to share oxygen between multiple patients (users would change the size of outlet nozzles to control flow rates), but these were frequently malfunctioning (e.g. missing plugs, blocked tubing) or used incorrectly, resulting in no gas flow to patients. 37,48 This led to the development of flowmeter assemblies, allowing individual titration of gas to multiple patients without changing plugs or connections (Gray et al.). 53,55 Power supplies were a major factor limiting the use of oxygen concentra-tors. 49,51,55,57 Uninterruptable power supply systems are useful if power outages are infrequent (e.g. less than daily) and brief (e.g. less than 20 minutes), but not if power failures are frequent or prolonged. 49 Solar power with battery storage is feasible, but requires experienced technicians and quality products. 26,30,31 Maintenance For users and technicians to provide reliable equipment maintenance and repair they need the capacity (knowledge and skills); opportunity (transport and access to spare parts); and motivation (role recognition and anticipated benefit). The Gambia provided a positive case study involving: well-trained biomedical engineers; a schedule for preventive maintenance; spare parts; electronic work orders (facilitating access to engineers); and a maintenance protocol for users. 48,49 They reported excellent long-term equipment function: 21/27 (78%) of oxygen concentrators working at 8-year follow-up; and minimal equipment down-time (5.2% of total cumulative hours in service). 48 They identified problems Malawi investigators used stepped-wedge methodology. The chronological data did not represent actual before-and-after intervention phases; the odds ratio was derived using unadjusted statistical methods. Notes: Studies were restricted to children aged < 5 years admitted to the paediatric ward. All studies used a before-and-after intervention study design. Systematic reviews Improving oxygen therapy for children Hamish Graham et al.
early (e.g. 37/53 (70%) repair needs identified and fixed during preventive maintenance visits) and showed that, while one third of concentrators needed repair each year, most repairs were cheap and simple (annual spare parts cost around 15 United States dollars per concentrator and required 50 person-hours). 48 Other projects with reasonable maintenance systems also reported good equipment function up to 3 years post-implementation (Gray et al.). 22,34,35,52,53 Projects where maintenance systems failed reported high equipment down-time and early equipment malfunction. 21,29,36 Many projects reported deficiencies in maintenance activities (e.g. site visits, spare parts, record-keeping and care by users), resulting in missed opportunities for early repair (Gray et al.). 35,37,46,48,49,53,60 Projects that integrated maintenance activities into existing (usually government) biomedical maintenance systems achieved success proportional to the quality of the existing system and its capacity to incorporate additional needs (Gray et al.). 34,35,46,53,60 Affordability Oxygen must be affordable for patients and hospitals. In user-pay hospitals oxygen was reported to be expensive for patients, contributing to delayed presentation, early discharge, and high patient debt burden (Gray et al.). 24,60 In the Lao People's Democratic Republic, oxygen from concentrators was pro-vided free of charge, thus reducing the cost of oxygen to patients and decreasing early discharge rates for unwell patients (Gray et al., Morpeth et al.). 46 In Nigeria, a neonatal unit used an oxygen insurance scheme to distribute the cost burden among all admitted patients and allow sustainable cost recovery. 24 Many projects reported that concentratorbased oxygen systems were relatively more cost-efficient than cylinderbased systems. 19,21,24,25,27,28,34,49,52,57,58,61 Demonstrated cost-efficiency can support the sustainability and future expansion of oxygen therapy, 21,24,26,49 but this may be limited if the decisionmaking body does not identify both financial and clinical benefits (Gray et al.). 34,57,58

Box 1. MEDLINE® search for publications on interventions to improve oxygen therapy systems in low-resource settings Search run on 10 August 2016
Improving oxygen therapy for children Hamish Graham et al.

Oxygen use
All the large-scale oxygen projects, and many single-site projects, aimed to improve how health-care workers used oxygen. However, while the overall use of oxygen increased, it was usually lower than expected based on admission and case-mix data, suggesting underuse of oxygen (Gray et al.). 35,37,40 The highest mean concentrator use (around 15-18 hours/day) was reported by two hospitals with high caseloads and close supervision by project doctors; 24,28 other projects reported a mean use of less than 6 hours per day (Morpeth et al.). 19,25,34,36,37,48,49,53 Post-implementation data from the Lao People's Democratic Republic, Malawi and Papua New Guinea showed persisting deficiencies in practice (< 50% of children with pneumonia received pulse oximetry; 22-80% of hypoxaemic children received oxygen) with great variation between hospitals (Morpeth et al.). 40,54 Hospitals with good pulse oximetry and oxygen practices achieved better clinical outcomes, with greater improvement if they started from a low baseline (Morpeth et al.). 53 Appropriate oxygen use was influenced by health-workers': (i) knowledge and skills; (ii) motivation; and (iii) work environment. 62

Knowledge and skills
Low knowledge and skills among health workers was almost universally reported, and many health workers had misconceptions and fears about oxygen therapy that affected their motivation (e.g. fear of the concentrator technology or belief that oxygen therapy caused death). Many projects provided initial training, and these generally showed improved knowledge and skills (Gray et al.). 38,45,52,53,61 High training coverage rates and in-service re-training were regarded as important in hospitals that faced chronic staff shortages and high staff turnover. 45,53 However, the proportion of staff trained correlated poorly with practice change or clinical outcomes, suggesting that training the right people in the right way may be more important than training more people. 38,52,61

Work environment
The physical and social environment could enable or inhibit good practice in oxygen therapy. For example, creation of high-dependency spaces within wards prioritized care for sick patients; endorsing protocols created new practice norms; wall-charts reminded users about oxygen use; logbooks facilitated regular equipment maintenance; flowmeter assemblies enabled oxygen delivery to multiple patients; pulse oximetry enabled identification of hypoxaemic children, helped involve families in care and overcame fears about oxygen; and oxygen analysers gave users confidence that the concentrator worked (Gray et al.). 28,35,46,48,49,53

Broader care practices
The projects in the Lao People's Democratic Republic, Malawi and Papua New Guinea embedded improved oxygen systems within broader measures to improve quality of care. All reported changes that extended beyond oxygenrelated care, such as improved staffing; reaching neglected populations; strengthening supply chains; and addressing other care deficits (Gray et al.). 46,53 While it was not possible to separate the effects of these quality of care interventions from the effects of the oxygen-related interventions, hospitals that improved the broader aspects of care typically reported the biggest improvements in outcomes (Gray et al.). 46,52,53,56

Supportive management
All the large-scale oxygen therapy projects required some degree of regional and national government support. In Egypt, the Lao People's Democratic Republic and Papua New Guinea effective relationships with government biomedical departments and the support of a multidisciplinary national oxygen team strongly facilitated equipment

Box 2. Theoretical themes identified to explain how interventions to improve oxygen therapy systems in low-resource settings might work
Theme 1 (oxygen access) recognizes that lack of oxygen in health-care facilities is a major barrier to care, and proposes that making oxygen therapy available will enable more children to be treated and with better outcomes. These projects typically emphasized the importance of quality oxygen equipment and effective maintenance programmes and some also addressed financial barriers to access. Theme 2 (oxygen use) recognizes that health-care workers may not use oxygen effectively, and proposes that building the capacity and motivation of health workers will enable children to be treated appropriately and have better outcomes. These projects typically emphasized training and supervision of clinical staff, including retraining and follow-up to ensure sustainability. Theme 3 (broader care practices) recognizes that broader issues of quality of care impact on clinical outcomes, and proposes that strengthening these processes will enable higher quality of care and better outcomes. These projects emphasized broader quality of care issues, such as supply of essential medications, clinical review and feedback and record-keeping. Theme 4 (supportive management) recognizes that managerial and political support is essential, and proposes that engagement with managers and policy-makers will support successful implementation and sustainability. These projects emphasized the engagement and responsibility of stakeholders at district, regional and national level in planning, implementation and evaluation. Theme 5 (hospital team) recognizes that hospital-level responsibility and action influences success, and proposes that enhancing the motivation of the hospital team will drive improvements in the use of oxygen and the care of equipment. This theme has not been overtly reported in previous oxygen therapy projects, but was frequently cited in interviews as a strong predictor of success by those involved.
Note: Potential theoretical themes were identified in the first stage of the realist review by a preliminary scan of the literature and interviews with key experts and stakeholders involved in oxygen projects. Themes were reviewed and refined during the extraction and analysis of data.

Systematic reviews
Improving oxygen therapy for children Hamish Graham et al.
maintenance (Gray et al.). 35,53 In Malawi, however, maintenance failed due to lack of support and supervision of the biomedical team. 46 Support from people within the Malawian health ministry enabled the programme to be sustain-ably incorporated into government child health programmes. 35.32 However, sustainability failed in Mongolia, partly due to the loss of key health ministry staff. In the Lao People's Democratic Republic and Papua New Guinea, ongoing train-ing has been supported by professional organizations and some regional health authorities, but sustainable programme implementation and scale-up has been limited by financial and practical constraints at the policy level (Gray et al.). 53

Hospital team
Investigators from many projects observed that ownership and acceptance of the project by hospital staff was a key determinant of success, enhancing: responsive problem-solving; equipment care and maintenance; clinical use of oxygen; continuity of knowledge and skills; and general quality of care (Gray et al.). 35,46,48,53 Limited data suggests that hospital-level ownership was greater when staff were convinced of the potential benefits of improved oxygen systems, and were actively involved in oxygen improvement activities as a multidisciplinary team. 24,28,52 Achieving hospital-level ownership may be more challenging for large-scale programmes that are initiated externally. A common approach was to find a responsible person at each hospital who would be the primary contact and local project leader. When it worked, this strategy created a local champion who had ready access to technical expertise and local staff, and ideally shared responsibility with a multidisciplinary team (Gray et al.). 28,46,48,53 This failed if the responsible person left their employment or lacked the power to effect change (Gray et al.). 46 While external agencies often brought funding and technical expertise to oxygen projects, studies reported that agencies could also be disruptive influences on hospital and health systems (Gray et al.). 29,46,52 Success depended on effective local participation (in planning, implementation and evaluation); clear communication between stakeholders (including priorities, roles and responsibilities); and ongoing support of local implementers (Gray et al.). 29,46,52

Discussion
The safe and effective provision of oxygen is a challenge for doctors, hospital administrators and government officials globally. Typically, some oxygen therapy equipment is available, but without the maintenance capacity to keep it functioning or the clinical expertise to use it effectively (Gray et al.). 36,40,46,50,56,63 While WHO guidelines offer advice on the technical specifications 9 and the clinical application of oxygen, 8 there has been little guidance on how to bring about changes in practice. In shifting the focus from what works to what happens, 64 realist review methods hold great promise for improving our understanding of how to bridge the implementation gap. 65 This is relevant for oxygen, and many other similarly simple, life-saving therapies that countries are struggling to scale up.
We identified a multilayered array of mechanisms that need to work together to achieve results. In broad terms, hospitals need good access to oxygen and effective use of oxygen, which should be facilitated by a broad quality improvement capacity, a strong managerial and policy support and multidisciplinary teamwork (Fig. 2). Within each domain, additional mechanisms are at play. It is the interaction of these processes, together with various contextual factors that determines the outcomes.
This framework is work in progress and it needs further testing. While there is substantial evidence to support some aspects of this framework, there is little evidence to support other aspects. Important future questions include: (i) the role of various policy-makers and managers in supporting the effective use of oxygen; (ii) the function of a multidisciplinary hospital oxygen team; (iii) the influence of broader care practices and potential role of quality improvement teams; (iv) the role of alternative energy sources in ensuring adequate power; and (v) how to integrate pulse oximetry and good oxygen practices more effectively into routine care. Those who are implementing an oxygen therapy system should not only evaluate the effectiveness of their programme, but test the mechanisms through which their programme works.
Our review had some limitations. First, realist reviews cannot explore all potential theories or all contextual influences. 11 We decided to include a wide range of interventions, but prioritized the exploration of theories that were most generalizable and applicable to low-resource hospitals providing general care for children. Our search criteria were therefore   11 We used published and unpublished evaluation reports, and interviews with study authors to clarify particular theories for which evidence was lacking. Different investigators, from different projects, may have provided different information. However, we do not believe this would have shifted the weight of evidence substantially, and we have documented where evidence is still lacking. Indeed, we believe that the validity of our findings was strengthened by exposing the data to critique from the implementers and other stakeholders of all the largescale projects identified.
In conclusion, our framework offers a practical, evidence-based approach to improving oxygen therapy in lowresource settings and may be relevant for other programmes involving the introduction of health technologies. It will assist practitioners and policy-makers to evaluate their current system(s), and create solutions that are appropriate for their context. It will assist implementers to understand and optimize their activities, providing both the flexibility and structure to adapt to different settings and respond to evolving needs. ■