Resilience in childhood vaccination: analysing delivery system responses to shocks in Lebanon

Introduction Despite rapidly growing academic and policy interest in health system resilience, the empirical literature on this topic remains small and focused on macrolevel effects arising from single shocks. To better understand health system responses to multiple shocks, we conducted an in-depth case study using qualitative system dynamics. We focused on routine childhood vaccination delivery in Lebanon in the context of at least three shocks overlapping to varying degrees in space and time: large-scale refugee arrivals from neighbouring Syria; COVID-19; and an economic crisis. Methods Semistructured interviews were performed with 38 stakeholders working at different levels in the system. Interview transcripts were analysed using purposive text analysis to generate individual stakeholder causal loop diagrams (CLDs) mapping out relationships between system variables contributing to changes in coverage for routine antigens over time. These were then combined using a stepwise process to produce an aggregated CLD. The aggregated CLD was validated using a reserve set of interview transcripts. Results Various system responses to shocks were identified, including demand promotion measures such as scaling-up community engagement activities and policy changes to reduce the cost of vaccination to service users, and supply side responses including donor funding mobilisation, diversification of service delivery models and cold chain strengthening. Some systemic changes were introduced—particularly in response to refugee arrivals—including task-shifting to nurse-led vaccine administration. Potentially transformative change was seen in the integration of private sector clinics to support vaccination delivery and depended on both demand side and supply side changes. Some resilience-promoting measures introduced following earlier shocks paradoxically increased vulnerability to later ones. Conclusion Flexibility in financing and human resource allocation appear key for system resilience regardless of the shock. System dynamics offers a promising method for ex ante modelling of ostensibly resilience-strengthening interventions under different shock scenarios, to identify—and safeguard against—unintended consequences.

Causal loop diagrams (CLDs) show cause and effect relationships between variables in a system.In a CLD, variables are linked by arrows which indicate both the direction of perceived effects, and -through the polarity (as indicated by a '+' or '-' sign)the nature of the relationship.A positive polarity indicates a reinforcing relationship between variables so that as one increases, the linked variable does in turn.A negative polarity indicates an opposing effect, so that a rise in the value of the starting variable leads to a reduction in the resultant variable.
The use of hashed lines across an arrow linking two variables indicates a delay in movement from one to the next.These delays can broadly be of two types: (i) information delays, which describe the time taken to assemble and interpret relevant information that enables a system change to be perceived or acted upon; and (ii) material delays, which describe the time taken for movement of materials affecting downstream system behaviours.Supply chain and logistics problems are good examples of material delays; in the case of vaccination delivery, one might consider the delay between disbursement of vaccine doses from a central storage facility to their arrival at a health service provider for administration, and the range of potential factors contributing to this.This visualisation shows the link between primary healthcare facility crowding and service user-perceived ability to socially distance in those facilities (in the context of COVID-19 spread).
For more complex behaviours, variables may be linked together to form loops, in which feedback behaviour occurs.These can be reinforcing or balancing.In a reinforcing loop, a series of variables may be linked together by positive polarities, so that a positive feedback loop is created.These kinds of loops can lead to rapid growth or decline in the variables of interest over time.A balancing loop, by contrast, tends towards equilibrium, and is seen when the relationships between variables within a loop even each other out.Further detail on the BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Figure 2 provides an illustrative example of a balancing feedback loop, based on the stem given in Figure 1.In this example, an increase in the perceived ability to socially distance in health facilities, reduces the perceived risk of contracting COVID-19, in turn promoting attendance at those facilities, with a resulting increase in vaccination uptake.However, there is a feedback effect because as attendance rises, so too does PHC crowding, which tends to reduce perceived ability to socially distance, ultimately discouraging further PHC attendance.Tomoaia-Cotisel A, Kim H, Allen SD, Blanchet K. Causal loop diagrams: a tool for visualizing emergent system behaviour.In: de Savigny D, Blanchet K, Adam T (eds.)

Figure 1
Figure1provides a simple illustration, showing the link between two variables used in this project: PHC crowding, and Perceived ability to socially distance in PHCs.In this example, as PHCs become more crowded, service users' confidence in being able to socially distance properly within those facilities reduces.There is also an information delay, in that it takes time for service users to become aware that local health facilities are crowded (unless they happen to be present in person).

Figure 1 .
Figure 1.Illustration of a negative polarity relationship between two variables using standard CLD notation.
can be found in a number of sources including Tomoaia-Cotisel et al(1), or for those with time and interest to pursue more detail, in Sterman (2) or Morecroft (3).

Figure 2 .
Figure 2. Example of a balancing feedback loop.
placed on this supplemental material which has been supplied by the author(s) supplemental material which has been supplied by the author(s)

Table 1 .
Key system responses identified by the main system level of action, and according to the category of response.BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) System level of action is given in the table rows; the category of response by the columns.Coloured dots indicate the points at which each response was observed (red dot = refugee crisis response; green dot = COVID-19; blue dot = the economic crisis).Bold letters indicate the kind of intervention implemented, using the WHO health system building blocks (F = financing; G = governance; I = information; M = medicines and technologies; P = people; S = service delivery; W = workforce)