Hospitals and healthcare facilities are vital assets to communities and play a key role in recovery from natural disasters. During emergencies, hospital units must provide uninterrupted critical services such as emergency care to the injured, laboratories, blood banks, ambulances, pharmacies and immunization services to prevent outbreaks of diseases (WHO, 2010). In recognition of the critical role played by hospitals in disasters, the Hyogo Framework for Action (UNISDR, 2005) and subsequent Sendai Framework (UNDRR, 2015), have as one priority the achievement of safe and resilient hospitals through structural, non-structural and functional risk prevention. This has resulted in major global initiatives for hospital safety and several guidelines have been issued for the design, assessment and strengthening of hospital buildings for different hazards (FEMA, 1997; FEMA, 2003; FEMA, 2007; PAHO, 2008; WHO, 2015). However, it is only relatively recently that tsunami design codes have been issued, e.g., FEMA 55 (FEMA, 2005), MLIT 2570 (MLIT, 2011), ASCE 7–16 Standard (ASCE, 2017a). These have not been implemented in the design of most healthcare facilities worldwide, and failures of hospitals in recent tsunami have caused extensive social and economic losses, (e.g. Kirsch et al., 2010; EEFIT, 2011). One means of disaster management for reducing life loss in tsunami is evacuation to sites outside the inundation zone or to upper levels in buildings considered strong enough to withstand the tsunami inundation (e.g. MHNI, 2015). Clearly, the vulnerable nature and reduced mobility of hospital patients makes evacuation difficult. Moreover, evacuation is only viable for locations that have tsunami warning systems in place and which are at a significant distance from the tsunami source.

Despite not being designed for tsunami, most hospitals are built to higher standards than normal residential buildings and present an enhanced resistance to natural hazards that may allow them to withstand small tsunami inundation without structural damage. However, hospital resilience relates not only to hospital building integrity, but also to maintaining hospital functionality. The latter depends heavily on the integrity of both non-structural elements and the lifelines supporting the hospital operation, such electricity, water and communications. The 2011 Tohoku tsunami presented several examples of hospitals that withstood the tsunami but had compromised functionality and ability to care for patients in the aftermath due to loss of lifelines and back-up systems in the tsunami inundation (EEFIT, 2011, EEFIT, 2013; ASCE, 2017b).

Hospitals can be considered as part of a network of healthcare provision, where only some parts of the network can be relied upon for the provision of any particular healthcare service, (e.g. not all hospitals have a trauma unit). As tsunami can affect large tracts of the coastline, they can damage several hospitals and/or supporting lifelines simultaneously. This not only disrupts the provision of healthcare locally but can result in the loss of particular healthcare services across large parts of the network, (e.g. if all hospitals with trauma units are affected over an extended region). Such scenarios result in affected people having to travel large distances and wait for excessive times to obtain specific treatments.

The inherent organisational complexity of hospitals, and the interactions and independencies of healthcare units makes the tsunami risk assessment of hospital services a challenging task. To date, several studies have investigated the performance of individual hospital buildings for different natural hazards using advanced engineering analysis, (e.g. Casarotti et al., 2009; Di Sarno et al., 2011). However, the use of advanced engineering analysis for the risk assessment of several hospitals is prohibitively expensive in terms of human and computational resources, as hospitals are typically composed of several buildings, built at different times and which do not follow a standard design. Furthermore, these studies rarely consider lifelines and back-up systems explicitly. As an alternative, several hospital safety indices (PAHO, 2008; WHO, 2015) and hospital safety checklists (WHO, 2008; WHO, 2010) have been proposed that offer rapid diagnostic tools for use by policy makers and hospital managers. These indices and checklists provide a qualitative estimate of the risk to hospitals from a set of hazards, i.e., natural and man-made hazards. The indicators can be applied to assess either single healthcare facilities or networks of hospitals, and generally account for the potential loss of critical infrastructure lifelines. These can be used to identify potential problem areas and for the prioritization of interventions to reduce the disaster risk to hospitals. However, these methods present two major shortcomings: 1) lack of quantitative approaches to support the assessment of the relative risk associated with the hospital facilities; and 2) little consideration of the nature of single hazards, (e.g. tsunami) and their interactions and interdependencies when impacting hospital infrastructure.

In order to improve both the safety and resilience of healthcare systems to tsunami, a simple but quantitative approach is required for assessing tsunami risk to healthcare services distributed across networks of hospitals. Such an approach needs to focus on healthcare service continuity, and go beyond hospital building integrity to consider the integrity of the lifelines and back-up systems that support the service provision and hospital functionality. This paper presents a new tsunami relative risk index ( T R R I ) developed to meet this need. Firstly, the components and calculation rationale for the T R R I are described. A survey form, specifically developed for collecting of field data on hospitals for the T R R I evaluation is also presented in the Appendix. In its current form T R R I is developed for hospital buildings of reinforced concrete construction, as these are the building types most commonly used worldwide for housing critical units, (e.g. Intensive Care Units). The T R R I is demonstrated through an application to 3 hospitals located along the southern coast of Sri Lanka (Galle, Matara and Hambantota Districts), which were surveyed by a team of researchers from UCL and University of Moratuwa. The T R R I is evaluated for three potential tsunami inundation events and is shown to be able to identify issues with both the buildings and functional aspects of hospital critical units. Three “what-if” intervention scenarios are selected and their effect on the T R R I is assessed. Through this exercise, it is shown that the T R R I can be used by decision makers to simply explore the effectiveness of individual and combined interventions in improving the tsunami resilience of healthcare provision across the hospital system.

Although the absence of numerical structural modeling to support the analysis can be seen as a limitation of this approach, the proposed relative risk index is based on objective engineering principles that are reflected in equations (and not merely expert opinions). The aim of using such an index is to be able to quickly assess a large portfolio of hospital facilities, identifying aspects of the facilities that require further detailed assessment, thus directing potential numerical modeling. For the case-study presented here, the inundation depths for the “what-if” intervention scenarios are based on limited onshore inundation scenarios based on the 2004 Indian Ocean Tsunami event, rather than on probabilistic data, since the latter is not available for Sri Lanka. Nevertheless, these scenarios give insight into the relative effectiveness of various mitigation measures that can be adopted by hospital administrators.

The proposed Tsunami Relative Risk Index ( T R R I ) aims to quantify the influence of the tsunami inundation on critical units, (e.g. Intensive Care Unit, Maternity Ward, etc) in individual hospitals, as well as the impact on service provision across a network of hospitals. The objective is to identify some of the drivers of risk to the hospital unit functionality, such that these can be prioritized for further investigation and intervention.

The proposed T R R I considers both the structural and functional attributes of hospital critical units, e.g., Intensity Care Unit, Maternity Ward, etc. The ability of a hospital critical unit to function in the aftermath of a tsunami depends on: 1) the stability of the structure where the hospital critical unit is located; 2) the integrity of non-structural elements relevant to the critical units, particularly the medical equipment that is required to ensure unit functionality; and 3) the functioning of the critical lifeline systems supporting unit functionality e.g., electric power, water supply, telecommunications, etc. Therefore, the proposed TRRI, for a hospital unit is defined as: T R R I = max ( R R I bldg ,   R R I funct ,   R R I bcs ) , (1) where R R I bldg considers the ability of the structural system to resist expected tsunami actions, R R I funct represents whether the location of the critical unit within the building puts it at high risk of loss of functionality under the expected tsunami inundation, and R R I bcs describes the risk of back-up critical systems to supporting lifelines being inundated. Each R R I component varies in value between 0 (no risk) and 1 (high risk). Each of these R R I components are further described in the following sections.

The TRRI Rapid Visual Survey (TRRI-RVS) form is developed to assist surveyors in assessing existing health facilities in terms of the integrity of hospital buildings, lifelines and back-up systems that support the service provision and hospital functionality. The TRRI-RVS form is presented in the Appendix (see Supplementary Material). The Rapid Visual Survey consists of two sections:

1) Hospital Profile (“Form H”). Through this form, surveyors collect general information about 1) the hospital location; 2) hospital type and hospital capacity, e.g., catchment population; 3) tsunami evacuation plans and disaster response plans; 4) hospital building locations within the healthcare facility; 5) location of critical hospital units within buildings, e.g., ICU, Labor Rooms, Maternity Wards, Pediatric Wards, Operating Theaters; and 6) presence and location of back-up supply systems.

The TRRI-RVS form is specifically developed for collecting the attributes of hospital surroundings, buildings, critical units, lifeline and back-up systems required to evaluate TRRI. This form is used in the survey of Sri Lankan hospitals used to test the TRRI in this paper.

Sri Lanka provides universal healthcare to its people through an established and robust healthcare system. Thanks to this, no major disease outbreaks occurred after the 2004 tsunami (Carballo et al., 2005), which hit two-thirds of the coastline affecting one million people. However, over 17% of all healthcare institutions were severely damaged, causing an estimated £40 M worth of losses (Komesaroff and Sundram, 2006). Over the last 15 years some of the affected health infrastructure of Sri Lanka has been re-built further inland, but some significant hospitals still lie within 2–3 km from the coast and are at potential threat from tsunami inundation. The Sri Lankan Ministry of Health (MoH) in collaboration with World Health Organization (WHO) has been working to strengthen the health sector for emergencies, through the development of a comprehensive national disaster management plan (WHO, 2015). However, this plan comprises capacity building in emergency management and health financing, and does not yet look at the structural, non-structural and functional performance of hospitals in natural hazards. Furthermore, as Sri Lanka is threatened by distal tsunami generated either at the Sunda trench or Makran Subduction zone, the main disaster management approach considered to date is the evacuation of hospitals (DPRD, 2015).

In this case study application, three hospitals in Galle, Matara and Hambantota Districts in Sri Lanka are selected for testing whether the TRRI can be used to 1) identify weaknesses in the systems supporting the functionality of critical units in individual hospitals, and 2) as a tool for use in prioritizing interventions for improved functional resilience across a series of hospitals.

The three hospitals selected are the District General Hospital (DGH) in Matara and the base Hospitals (BH) in Balapitiya and Tangalle. These are chosen as they are key hospitals for the Southern Province, geographically distributed across the Province (Figure 4) and all located within 400 m from the coast (base Hospitals) or a litte further (approx. 600 m) but near a waterway that discharges into the sea (DGH Matara). The case study application focuses on the five critical units that were indicated as the most important in the case of a disaster by the Disaster Preparedness and Response Division (DPRD) of the Sri Lankan Ministry of Health, Nutrition and Indigenous Medicine. These are: 1) Intensive Care Units (ICU); 2) Operating Theaters (OT); 3) Labor Rooms (LR); 4) Maternity Wards (MW); and 5) Pediatric Wards (PW). In the three hospitals, 19 buildings were found to house these critical units, and were surveyed by a joint team from UCL and University of Moratuwa in April 2019 using the form described in TRRI Rapid Visual Survey (RVS) Form.

Thirteen of the buildings are reinforced concrete moment resisting frame structures of 2–4 storeys. These house 85% of all the critical units in these three hospitals. The remaining six buildings are one-storey load-bearing unreinforced masonry (URM) structures (Figure 5). These structures are highly vulnerable to tsunami and would not be expected to be in an operational state following tsunami inundation. Hence, this assessment concentrates on the assessment of the 22 critical units housed in the RC buildings. The survey of these buildings highlighted that most of the critical units are located at the ground floor and are therefore at high risk from damage if the tsunami inundation reaches the building. The soil type at each hospital is determined as non-cohesive from observational and borehole data analysis. Hence all buildings are susceptible to scour in this case study application. None of the buildings assessed were located near ports and harbors, and are therefore not exposed to impact from containers, ships or barges. Consequently, the assumption of logs as debris is appropriate for this case study.

The surveys showed the HVAC to be a local system of air conditioning units attached to the walls of critical units. Hence, they will continue to function if the critical unit is not inundated. The location of TLC systems is assumed to be in the hospital administrative offices. This is because Hospital Directors and administrative staff typically have access to the emergency systems for communicating with the national and district-level healthcare networks. Where back-up systems were not recorded during the field survey it is assumed they are missing. As this is detrimental to functional resilience, these back-up systems are still included in the calculation of RRI _bcs and contribute to increasing its value. For example, no fire alarms, extinguishers or other fire protection systems were observed in any of the assessed buildings, hence a value of FP = 1 is applied for all buildings within the RRI _bcs calculation.

Table 4 presents the values of TRRI calculated for the five critical units of the three case-study hospitals, for the three hazard scenarios presented in Hazard Scenarios. Under Hazard Level 1, none of the buildings containing critical units in BH Balapitiya and BH Tangalle are subjected to tsunami inundation. Despite this, the R R I bcs values for these hospitals are non-zero due to their both not having any fire protection system, and BH Tangalle missing power and water back-up systems. For DGH Matara, the values of R R I bldg indicate that only building M15 would be likely to collapse due to scour ( R R I scour = 1 , see Table 5), with the other buildings not suffering major damage, (i.e. R R I bldg ≤ 0.5 ). Despite the good building performance, five of the critical units would be directly inundated ( R R I funct = 1 ) , and four more critical units would likely be non-functional due to compromised back-up systems ( R R I bcs = 1 ) . The latter is due to the main back-up systems in this hospital being inundated. The consequence is that under this hazard scenario (and also for Hazard Levels 2 and 3), DGH Matara is predicted to lose functionality in all its critical units. Across the network of these three hospitals, this would mean a reduction of 40–45% in the number of ICU and MW units, and of 50% in the number of LR and OT units. Loss of critical unit functionality at DGH Matara would put particular stress on BH Tangalle, which is the closest hospital to it, and which has only two ICU units overall (only one in an RC building) and no Operating Theater.

Under Hazard Level 2, BH Balapitiya remains outside the inundation zone, but building T9 of BH Tangalle is subjected to a small inundation of 0.29 m depth. This inundation is insufficient to cause structural damage in this building but does compromise the functionality of one of the Maternity Wards, as this is located at the ground floor of T9. Moreover, all other critical units in BH Tangalle are seen to be at significant risk of functionality loss from damaged back-up systems. Hazard Level 2 imposes a larger inundation depth at DGH Matara, which results in three predicted building collapses ( R R I bldg = 1.0 ) . Through analysis of the components of R R I bldg (see Table 5), the risk of structural failure from hydrodynamic loading is significantly higher than in Hazard Level 1, but overall building failures are dominated by the effects of scour around the foundations. With almost all the critical units in both BH Tangalle and DGH Matara predicted to be non-operational (see Table 4), Hazard Level 2 sees a reduction across the three hospitals of 55% in the number of ICU units, 50% in the number of LR and OT units, and 80% in number of MW units.

When subjected to Hazard Level 3, all critical hospital units would likely be non-operational. As listed in Table 5, all hospital units in DGH Matara are located within buildings at significant risk of structural damage and severe scouring at the foundations. At BH Balapitiya, although power, water supply and medical gases would continue to function ( R R I bcs = 0.5 ) (Table 4), two buildings (B7 and B11) would be at high risk of collapse due to effects of scour and debris impact (Table 5). This would make two ICUs and one MW non-operational, despite their being located on building storeys that would not be inundated by Hazard Level 3. For 64% of the units across the three hospitals R R I bcs = 1 , since the backup systems would be compromised. At BH Tangalle, the lack of power and water supply combined with damage to the rest of the back-up systems, results in R R I bcs = 1 for all units. If this can be prevented, BH Tangalle would be able to operate 50% of its the Maternity and Pediatric Wards (since buildings T1 and T9 have R R I bldg = 0   and   0.6 , respectively, and their first floors have R R I funct = 0 even for Hazard Level 3–see Table 4).

The results of the analysis of TRRI for the three hospitals and Hazard Levels shows a high vulnerability of back-up systems and critical units under low levels of tsunami inundation. This is caused by most being located on the ground floor of inundated buildings (see Table 4). These two components of TRRI are seen to dominate whether or not critical units will be operational after a “small to moderate” tsunami event (Hazard Levels 1 and 2). Note that TRRI = 1.0 for nearly half of the units (45% of the total) at Hazard Level 2, although R R I bldg = 1.0 only for 18% of them. Hence, re-positioning critical units and back-up systems to higher floors within the surveyed buildings would improve the functional resilience of the hospitals. Building failure plays an increasing role in the critical unit operationality for “moderate to high” tsunami events (Hazard Levels 2 and 3). At Hazard Level 3, all 22 units have TRRI = 1.0, of which 13 units (59%) also have R R I bldg = 1.0 . In particular scour of foundations can precipitate building failure. Protection against scour would require the installation of piles or deeper foundations. This is more appropriate as a design improvement for future hospital buildings, since this can be disruptive and expensive as a retrofit intervention.

Given the findings in Results of the Assessment of Critical Units for Sri Lankan Hospitals, this section presents a comparison of the effectiveness of three possible interventions in reducing the immediate loss of operationality of critical units after a tsunami. The intervention effectiveness is examined by running “what-if” scenarios, wherein the intervention is applied to all buildings and TRRI is recalculated. The effectiveness of the intervention on each critical unit type is represented as the ratio between the number of functional units for the intervention and baseline scenarios (Note: the baseline is the no-intervention scenario). The “what-if” scenarios considered are: •

What-if 1 (WI1) consists in the relocation of back-up systems to places that are not affected by the tsunami inundation, e.g., by either relocating or elevating the system to be outside the inundation zone. Within this scenario, any missing back-up system, other than Fire Protection and HVAC (as these are co-located with the critical unit) are installed.

Table 6 presents the TRRI resulting from implementation of the three “what-if” scenarios and the baseline (no intervention) scenario for the three Hazard Levels. Table 7 summarizes the effectiveness of each “what-if” scenario in increasing the number of operational critical units after a tsunami, as compared to the baseline scenario. In Table 7, the effectiveness of the “what-if” scenario, indicated as E WI , is calculated for each critical unit type, as follows: E W I = n ou,   WI − n ou,   BL n u , (25) where n u is the total number of units (for each type), n ou , WI is the number of operational units in the “what-if” scenario, and n ou , BL is the number of operational units for the baseline scenario.

From Tables 6 and 7 it is observed that moving the back-up systems to a safe location (WI1) significantly improves the number of operational MW, OT and PW available after tsunami for all Hazard Levels, but is not effective in improving the number of operational ICU and LR units with respect to the baseline for tsunami above Hazard Level 1. This is because many critical units remain vulnerable to direct tsunami inundation.

Implementation of WI2 provides no/little improvement over the baseline scenario for Hazard Levels 1 and 2, as the failure of back-up systems in DGH Matara and BH Tangalle compromise their critical unit operationality and BH Balapitiya is not inundated at these Hazard Levels. However, for Hazard Level 3, despite inundation of BH Balapitiya, some of the back-up systems are not compromised and by elevating the critical units their risk of direct inundation is reduced and their operationality maintained.

An increased effectiveness is observed for What-If scenario 3, as compared to either WI1 or WI2 individually. The combined intervention on back-up systems and critical units is more beneficial than the sum of their individual effects. This is because in WI3 any missing back-up systems are added to the hospital buildings, and the HVAC and Fire Protection systems are moved to upper levels with the critical units, thus joining the other back-up systems in being in a safe location. This results in RRI _bcs values close to zero, which when combined with the reduced risk of critical unit inundation, results in 95%, 82%, and 64% of all critical units being operational under Hazard Levels, 1, 2, and 3, respectively. It is highlighted that even in WI3, ICU and OT remain at significant risk from tsunami of Hazard Level 3, with only one quarter of the units predicted to remain operational. To further increase their tsunami resilience, interventions would be needed on the buildings that house these critical units, in order to improve their structural and foundation systems. The TRRI analysis prioritizes buildings M1, M15, and M27 in DGH Matara for such interventions, as these are predicted to suffer heavy damage and/or collapse under the tsunami hazard scenarios, even though the risk to back-up systems and critical units can be reduced through WI3.

The suggested interventions are not based on financial considerations or other constraints, and are applied to all three hospitals. However, it is clear that the TRRI and proposed efficiency measure ( E WI ) can be adopted for other What-If scenarios that could apply more targeted or different interventions on single hospitals or buildings to optimize the cost-to-benefit. The advantage of the TRRI is that such interventions can be explored across single or multiple hospitals in a manner that is not computationally expensive and does not require high levels of technical expertise.

This paper presents a new tsunami relative risk index ( T R R I ) for the assessment of risk to critical units in hospitals exposed to tsunami inundation. The TRRI is a quantitative index that considers tsunami risk to 1) the hospital buildings housing critical units, with tsunami hydrodynamic loading, debris impact and scour considered, 2) the critical units themselves and 3) the critical back-up systems that support the functioning of critical units. Each component of tsunami risk is evaluated on a scale of 0 (no risk) to 1 (high risk), and the overall risk to the critical unit is taken as the highest value from the three components. A methodology is provided for the simple evaluation of the tsunami risk indices for each component that draws upon engineering principles and practice, physical interpretation of tsunami risk and expert elicitation. The TRRI approach is tested for a case study of three hospitals in Sri Lanka, wherein the TRRI is used to assess the number of critical units (that are housed in reinforced concrete buildings) remaining operational after tsunami inundations of three intensities. It is demonstrated that the TRRI approach allows the identification of the drivers of loss of operationality of critical units under the different hazard scenarios. The TRRI analysis for the three hospitals show a high functional vulnerability of back-up systems and critical units under low levels of tsunami inundation. These findings can inform decisions to be made as to interventions for improving the operational resilience of critical units within a single hospital complex, as well as across a network of hospitals to ensure health service provision. The latter is demonstrated by conducting a series of “what-if” scenarios for different interventions on the case study hospital network and re-calculating the TRRI values for each critical unit. Comparison of the number of critical units predicted to be functional after a tsunami under the baseline scenario, (i.e. no intervention) and the different “what-if” scenarios, allows the identification of individual and combined interventions in improving the tsunami resilience of healthcare provision across the hospital system. For the three hospitals in Sri Lanka, relocating back-up systems and units to safe locations would be an effective intervention; however, under large tsunami events the hospital buildings and their foundations are predicted to suffer heavy damage and/or collapse.