Attacks on healthcare in conflict-affected countries: a comparison of temporal trends in ongoing conflicts in Lebanon, Myanmar, occupied Palestinian territory, Sudan and Ukraine using WHO SSA and SHCC data, 2018–2024

Abstract

Background

The Geneva Conventions form the core of International Humanitarian Law (IHL), safeguarding healthcare and protecting civilians from the brutality of war. Unfortunately, these conventions are often disregarded. Attacks on healthcare have devastating effects on healthcare systems, and it is therefore vital to document such attacks and detect possible temporal patterns. This study aims to assess temporal trends in attacks on healthcare in five conflict-affected countries: Lebanon, Myanmar, occupied Palestinian territory (oPt), Sudan and Ukraine.

Methods

This study used two publicly available databases: the World Health Organization Surveillance System for Attacks on Health Care (WHO SSA) and the Insecurity Insight Safeguarding Health in Conflict Coalition (SHCC). Start dates and key events were determined for each conflict based on grey literature searches. From the start dates onward, data on attacks on healthcare were collected. The data collection ended on December 31, 2024. Statistical analysis entailed chi-square tests for temporal trends.

Results

The WHO SSA and SHCC database reported a total of 4,289 and 5,454 attacks, respectively, in the five investigated conflict-affected countries. For all conflict-affected countries except Lebanon, there were significant differences between the databases regarding the reported number of attacks. Temporal trend analyses revealed that, in Myanmar, oPt, Sudan and Ukraine, the highest number of attacks occurred during the 0–2 month period. In Lebanon, the highest number of attacks was observed in the 9-11-month period. All peaks in the number of attacks were associated with either the immediate or early phase of the conflict or with major conflict escalations.

Conclusions

Temporal trend analyses of five ongoing armed conflicts revealed that spikes in attacks on healthcare were either associated with the immediate or early phases of the conflict or with major conflict escalations. Although major differences exist between the WHO SSA and SHCC database, particularly regarding the reported number of attacks, the observed patterns were largely similar.

Background

In times of war or armed conflict, there will always be people and material objects that fall victim to the brutality of violence. In order to limit the barbarism of war and safeguard healthcare in conflict, the Geneva Conventions and additional protocols were established. The Geneva Conventions form the core of international humanitarian law (IHL). These treaties state that people who are not, or are no longer, directly or actively participating in hostilities, are protected during armed conflicts [1, 2]. This includes healthcare workers and their medical facilities. Unless medical units commit harmful acts to either party in a conflict, they are to be protected in all circumstances. Otherwise, attacks on healthcare are grave breaches of IHL and the Geneva Conventions in particular [3–6]. Despite these international conventions, the protected status of healthcare is often disregarded. In fact, it is suspected that healthcare facilities, such as hospitals and ambulances, are increasingly subject to deliberate or indiscriminate attacks.

For example, Kunichoff et al. (2024) investigated whether the damage to and destruction of hospitals in the Gaza Strip sustained by 2,000 lbs bombs were deemed collateral damage or if it could imply deliberate targeting. Combining available geospatial data for the identification of hospital sites with satellite imagery investigations from CNN and New York Times, it was estimated that 84% of all hospitals were within the range of damage and injury, and 25% of all hospitals were within the ‘lethal’ range. Furthermore, Hayes Wong et al. (2018) performed a literature review on ambulance attacks in Syria between 2011 and 2018, including a descriptive secondary data analysis on individual attacks that were reported to the Syrian Network for Human Rights. This secondary data analysis suggested that attacks on ambulances are relatively common. Furthermore, ‘double-tap’ attacks were frequently observed. This involves a tactic in which a second strike follows closely after an initial attack, deliberately targeting first responders [7, 8]. Another study investigated the relationship between facility size and damage sustained during Russia’s invasion in Mariupol, Ukraine [9]. 77% of the medical facilities in Mariupol had sustained damage during the conflict. Facility size was not associated with damage, suggesting that attacks on medical facilities are not random but instead may have been the result of intentional targeting.

Attacks on hospitals have severe consequences. Short-term effects range from damage to facilities and nearby infrastructure, thus leading to decreased access and quality of care, increased costs, higher workloads for healthcare workers and a drop in morale [10]. Long-term effects are shown to be just as serious. Disruption of healthcare services has a significant effect on patients with chronic conditions and those who require specific care [11, 12]. Furthermore, attacks on healthcare are associated with increased spread and reduced control of infectious diseases, reduced healthcare access and considerable loss of the healthcare workforce [10–13]. Moreover, such attacks have been shown to precede conflict escalation in affected areas and are associated with forced displacement of the local population [12–14].

Temporal trends in attacks on healthcare facilities have occasionally been studied in individual conflicts. A study dissecting attacks on Ukrainian healthcare facilities during the first year of the conflict demonstrated a cluster of attacks in the first three months of the conflict [15]. Another study assessed temporal trends and attack types in the occupied Palestinian territory (oPt). In the first four months, there were multiple clusters of attacks with high peaks, after which the peaks decreased. These studies suggest that attacks on healthcare are particularly prevalent during the onset of a conflict, which warrants further research [16].

In this study, we aimed to examine temporal trends in attacks on healthcare across five conflict settings, Ukraine, the oPt, Lebanon, Myanmar and Sudan, using publicly available data to compare patterns of violence against health services in these conflicts.

Methods

Study design

This was a retrospective observational study using publicly available data from the World Health Organization Surveillance System for Attacks on Health Care (WHO SSA) and the Insecurity Insight Safeguarding Health in Conflict Coalition (SHCC) databases, which include attacks on healthcare as of 2018 and 2016, respectively [17, 18]. Five conflict-affected countries were selected for this study: Ukraine, the oPt, Lebanon, Myanmar and Sudan. These conflict-affected countries were selected because the respective conflicts were considered recent and major. Furthermore, the databases used for this study included evidence of attacks on healthcare facilities during these conflicts in conflict-affected countries.

To ensure consistency, the analysis was conducted at the conflict area level rather than at the specific armed conflict level, as some conflicts span multiple geographic regions, complicating attribution and analysis.

The analysis for each conflict-affected country starts from a start date (or set point) in the history of the conflict. All conflict analyses ended on December 31, 2024. The start dates were determined for each conflict-affected country and either corresponded to the actual start of the conflict or to a major escalation of an ongoing conflict based on grey literature searches. Further details on the start dates per conflict-affected country are provided below.

The available data were directly exported to and edited in Microsoft Excel (Microsoft Corporation, Redmond, Washington, United States of America). Editing of the existing databases was performed to compare the two databases and to conduct statistical analysis. Statistical analysis was conducted in SPSS version 28 (IBM Corp., Armonk, New York, United States of America).

Description of databases

The WHO SSA database was established in 2017 and documents which aspects of healthcare were targeted, including healthcare facilities, transportation, personnel, supplies and assets, warehouse storage and patients. Further information was provided regarding attacks on healthcare facilities, differentiating between primary, secondary, tertiary or other facility types. The database also includes abductions, arrests and detentions of healthcare workers and patients and documents the number of people injured and killed in these attacks.

The SHCC database was established in 2014 and focuses on healthcare facilities, healthcare transportation facilities and healthcare workers. It categorizes impacts on facilities as follows: forceful entry into health facilities, occupation of health facilities and vicinity of health facilities affected. Healthcare transportation was categorized as destroyed, damaged, and stolen or hijacked. The database includes impacts on healthcare workers: killed, injured, kidnapped, arrested, threatened, assaulted and sexually assaulted. Other variables were health supplies (looting, theft, robbery or burglary) and access denied or obstructed.

Determination of the starting date per conflict-affected country

A grey literature search was performed to investigate each ongoing conflict. In this search, the set points were determined for each conflict, as were the key moments within each conflict. These key moments involved turning points or escalations that significantly altered the course of the conflict and were determined by two of the authors (JU and DB). The authors reached consensus on all key moments. These moments were later used to assess possible correlations with temporal trends of attacks in the databases.

Myanmar

Since 1948, Myanmar has been involved in armed conflict, when the country, then known as Burma, gained independence from the United Kingdom [19–21]. The conflict has been mostly based on ethnicity, with armed ethnic organizations fighting the armed forces of Myanmar for sovereignty. In 2021, a military coup by commander-in-chief Min Aung Hlaing deposed the civilian government, setting off widespread protests and escalating insurgencies. The start date for this non-international armed conflict was February 1, 2021, when the military coup began.

Ukraine

Ukraine and Russia have been in conflict since Russia annexed Crimea in 2014 [22, 23]. Russia launched a full-scale invasion of Ukraine on February 24, 2022 and began occupying larger parts of the country. This started the greatest conflict in Europe after World War II. Therefore, February 24, 2022, was chosen as the start date for this international armed conflict.

Sudan

Sudan has endured chronic instability marked by prolonged military rule, 20 coup attempts, two civil wars, and the Darfur genocide since gaining independence in 1956 [24–26]. In 2019, another coup took place, carried out by the Sudanese Armed Forces (SAF) and the Rapid Support Forces (RSF). Their collaboration lead to an ongoing fight for power between the leaders of the groups, and escalated on April 15, 2023, marking this start of this most recent non-international armed conflict in Sudan.

Occupied Palestinian territory (oPt)

The conflict between Israel and Palestine is an ongoing political and military conflict over sovereignty and land within the territory of the former Mandatory Palestine. Key aspects of the conflict include the Israeli occupation of the West Bank and the Gaza Strip, the status of Jerusalem and the expansion of Israeli settlements [27, 28]. On October 7, 2023, Hamas-led militant groups launched a surprise attack on southern Israel, killing more than 1,200 Israeli civilians and military personnel. Furthermore, approximately 250 people were taken hostage into Gaza. In response, Israel’s government declared war on Hamas and conducted an extensive aerial bombardment campaign on Gaza. This was followed by a large-scale ground invasion with the stated goal of destroying Hamas, freeing hostages, and controlling security in Gaza afterwards. The start date for this international armed conflict was defined as October 7, 2023.

Lebanon

The Israeli–Lebanese conflict is a long-running conflict that involves Israel, Lebanon-based paramilitary groups, and, at times, Syria [29–32]. During the Lebanese Civil War, the conflict peaked. Israel invaded the country in 1978 and in 1982 in response to Palestinian attacks from Lebanon. It occupied southern Lebanon until 2000 while fighting a guerrilla conflict against Shia paramilitaries. After Israel withdrew, Hezbollah attacks triggered the Lebanon War in 2006. A new phase of the conflict started in 2023, which ultimately led to the Israeli invasion of Lebanon on September 30, 2024. The start date of this international armed conflict was October 8, 2023, when Hezbollah launched a missile attack on positions in Shebaa Farms occupied by Israel during the 2023 Gaza war.

Definitions

An attack was defined according to the WHO definition, which is used by both the WHO SSA and the SHCC databases: “any act of verbal or physical violence, threat of violence or other psychological violence, obstruction that interferes with the availability, access and delivery of curative and/or preventative health services”.

Categories of types of attack

WHO SSA

The WHO SSA database categorizes different types of attacks [33]. These are defined as displayed in Table 1:

Table 1.

Definition of attack types used in the WHO SSA database

Type of attack	Definition
Abduction, arrest or detention	The unlawful removal, seizure, capture, apprehension, taking or enforced disappearance of a person, either temporarily or permanently; a seizure or forcible restraint, an exercise of power to deprive a person of his or her liberty; the condition of being held in confinement and deprived of personal liberty.
Armed or violent search of health care personnel, facility or transport	Examination of a person’s body or property, or of a health care facility or transport through use of physical violence, weapons or of other means of coercion.
Assault	Violence inflicted from one person to another person who does not involve individual weapons or heavy weapons.
Obstruction to health care delivery (e.g. physical, administrative or legal)	Any act resulting in blocking or preventing access to health care.
Psychological violence/threat of violence/intimidation	Emotional abuse, such as insults, belittling, humiliation, intimidation or threats of harm. The explicit declaration of a plan, intention or determination to inflict harm, whether physical or psychological.
Removal of health care assets	This can include looting, robbery or theft of health care assets. Looting refers to stealing of goods, typically during a war or riot. Theft is defined as the unauthorized taking of property from a person through the offender’s use of physical force or threat of physical force.
Setting fire	Arson; the act of malicious burning.
Violence with heavy weapons	Violence with a weapon that requires more than one person to use, such as knives, bricks, clubs, guns, grenades and improvised explosive devices (IED)

.

Insecurity insight SHCC

The SHCC database focuses on the types of weapons that were used instead of attack types. These included the following categories: aerial bomb (distinguished between drone, missile and plane), arson, artillery, cluster bomb, firearms, fist and foot, hand grenade, knife, improvised explosive device (IED) (distinguished between radio controlled, remote controlled, suicide-vehicle born, vehicle borne or unspecified), mine, missile, mortar, rocket propelled grenade, shelling, stones, sticks and gravel, tasers, live and rubber bullets, unexploded ordnance, unarmed perpetrator, unspecified explosive, and other weapons. The SHCC database also includes the date of the attack.

Ethical approval

This study involved the analysis of publicly available databases. No patients, individual participants or trials including animals were included in this study, and there was no need for METC or DEC approval. Local approval was obtained from the board of directors of VieCuri Medical Center, Venlo, the Netherlands (#2025.11).

Statistical analysis

To analyze the study variables, each dataset was organized by conflict-affected country and date of attack. Each conflict-affected country was assigned a start date as described above. From that date, time periods of three months were created so that the selected conflicts could be compared. Chi-square tests were subsequently applied to evaluate trends of attacks over time, with a significance level of p < 0.05.

The statistical analyses were conducted via SPSS version 28 (IBM Corp., Armonk, New York, United States of America), Microsoft Excel (Microsoft Corporation, Redmond, Washington, United States of America) and an online calculator (GraphPad, San Diego, California, United States of America) [34].

Results

For all the studied conflict-affected countries in the investigated time periods, the WHO SSA and SHCC databases included a total of 4,289 and 5,454 attacks, respectively.

WHO SSA database

The WHO SSA database reported 2,968 injuries and 1,695 fatalities. Among the total 4,289 attacks, healthcare facilities were involved in 2,882 cases, transport in 1,135 cases, personnel in 1,838 cases, supplies and assets in 713 cases, warehouse storage in 122 cases and patients in 1,236 cases.

The exact facility type was unknown in 1,797 attacks. For the remaining 2,499 attacks, 1,020 attacks affected primary healthcare facilities, 728 affected secondary healthcare facilities, 165 affected tertiary healthcare facilities and 585 affected other healthcare facilities, such as pharmacies, mobile clinics and unspecified facilities.

In total, 1,695 and 2,968 people were killed and injured, respectively. Healthcare workers were further affected by 172 abductions, 525 arrests and 331 detainments. Furthermore, 35 patients were abducted, 60 were arrested, and 40 were detained.

SHCC database

Among the 5,454 attacks in the SHCC database, there were 174 cases in which the healthcare facility was destroyed, 1,531 cases of damage, 613 forceful entries, 344 occupations of healthcare facilities and 424 cases in which the vicinity of the healthcare facility was affected.

Among the attacks involving healthcare transportation, 90 involved destroyed transportation, 431 involved damaged transportation, and 199 involved stolen or hijacked transportation.

Among healthcare workers, 1,107 were killed, 651 were injured, 886 were threatened, 87 were kidnapped, 1,333 were arrested, 136 were assaulted, and 30 were sexually assaulted.

There were 607 reports of healthcare access denial or obstruction and 260 instances of looting, theft, robbery or burglary of health supplies.

Table 2 shows the absolute numbers of incidents reported in the databases.

Table 2.

Description of the WHO SSA and SHCC database

	Lebanon	Myanmar	Occupied Palestinian territory	Sudan	Ukraine	Total numbers
WHO SSA
Total number of attacks	160	443	1,338	136	2,212	4,289
Healthcare system
Facilities Transport Personnel Supplies and assets Warehouse storage Patients	45 104 111 38 11 6	188 59 249 82 49 121	688 564 1,139 47 4 958	96 16 60 58 30 42	1,865 392 279 488 28 109	2,882 1,135 1,838 713 122 1,236
Type of facilities	27	218	3	99	2,152	2,499
Primary Secondary Tertiary Other Unknown	8 16 0 3 133	103 50 18 47 225	0 1 1 1 1,335	12 51 16 20 37	897 610 130 514 60	1,020 728 165 585 1,797
Killed	241	108	903	238	205	1,695
Injured	295	289	1,472	214	698	2,968
Healthcare workers
Abduction Arrest Detention	0 0 0	127 379 12	27 143 310	16 3 6	2 0 3	172 525 331
Patients
Abduction Arrest Detention	0 0 0	13 15 0	7 45 40	0 0 0	15 0 0	35 60 40
SHCC
Total number of attacks	153	1,492	1,678	461	1.670	5,454
Healthcare facilities
Destroyed Damaged Forceful entry Occupation Affected vicinity	1 57 0 0 17	72 249 371 190 51	28 225 73 13 283	22 100 59 50 25	101 900 110 91 48	174 1,531 613 344 424
Transportation
Destroyed Damaged Stolen or hijacked	4 26 0	38 60 72	15 102 1	2 3 50	31 240 65	90 431 188
Healthcare workers
Killed Injured Threatened Kidnapped Arrested Assaulted Sexually assaulted	111 89 4 0 0 0 0	126 126 663 35 918 56 2	511 161 124 0 357 23 26	108 78 39 43 24 49 2	251 197 56 9 34 8 0	1,107 651 886 87 1,333 136 30
Denied or obstructed access to healthcare	3	145	388	18	53	607
Looting/theft/robbery/burglary of health supplies	0	58	5	74	123	260

Temporal trends

Myanmar

The highest number of attacks in this conflict-affected country occurred in the 0–2 month period of both databases (WHO SSA: 183 reported attacks; SHCC 212). The lowest number of attacks was observed in the 42–44 month period of both databases (WHO SSA: 1 reported attack, SHCC 43).

The temporal trends were analyzed with the chi-square test, resulting in a p value of <0.001 for the WHO SSA database and <0.001 for the SHCC database, indicating a difference in the number of events between 3-month periods.

After the initial peaks of the conflict, the graphs in Fig 1 show multiple smaller peaks in the number of attacks. The first distinguishable peak occurred at approximately 10 months, and there was a cluster of peaks between 23 and 33 months.

Fig. 1.

Temporal trends of attacks on healthcare in Myanmar, WHO SSA and SHCC database. 10 months (late 2021): entire villages were destroyed, causing a massacre of both civilians and opposition fighters. 23 months (late 2022): beginning 2023, resistance forces overcame weaknesses in command and control

Ukraine

The highest number of attacks was reported in the 0–2 month period for both databases (WHO SSA: 89 reported attacks; SHCC: 150). The lowest number of attacks was observed in the 33–35-month period for the WHO SSA database (n = 18) and in the 30–32-month period in the SHCC database (n = 13).

Temporal trends in both databases were analyzed with the chi-square test, resulting in p values of < 0.0001 for the WHO SSA database and < 0.001 for the SHCC database, indicating a difference in the number of events between the 3-month periods.

Figure 2 shows that the highest number of attacks took place in the 0–2 month period after the start of the conflict. After that initial peak, significant events in the history of the conflict do not appear to align with noteworthy peaks in the graphs.

Fig. 2.

Temporal trends of attacks on healthcare in Ukraine, WHO SSA and SHCC database. 0-2 months: Russian forces focused on eastern and southern Ukraine, after failing to take Kyiv. 7 months (September 21, 2022): Russia declared a partial mobilization and committed to defending. 14 months (June 2023): Ukraine launched counteroffensive. 18-21 months (winter 2023): Stalemate of conflict. 26 months (May 2024): Russia opened new front in northeastern Ukraine

Sudan

The highest number of attacks was observed in the 0–2 month period of both databases (WHO SSA: 50 reported attacks; SHCC: 172). The lowest number of attacks occurred in the 9–11 month period of both databases (WHO SSA: 1; SHCC 44).

Temporal trends were analyzed with the chi-square test, resulting in a p value of < 0.001 for the WHO SSA database and < 0.001 for the SHCC database, indicating a difference in the number of events between the 3-month periods.

In Figure 3, the highest peak in the graphs reflects the first months of the conflict. Afterwards, a distinguishable peak was observed at the 8-month mark, and there was an increase in attacks after 12 months.

Fig. 3.

Temporal trends of attacks on healthcare in Sudan, the WHO SSA and the SHCC database. 12 months (March 8, 2024): attempted ceasefire. 17 months (September 26, 2024): The Sudanese army carried out raids against the Rapid Support Forces in the capital in its biggest assault in months

Occupied Palestinian territory

The highest number of attacks occurred in the 0–2 month period of both databases (WHO SSA: 653 reported attacks; SHCC: 918). The lowest number of attacks was reported in the 6–8 month period of the WHO SSA database (136) and in the 9–11 month period of the SHCC database (130).

Temporal trend analyses via the chi-square test revealed a p value of < 0.001 for the WHO SSA database and < 0.001 for the SHCC database, indicating a difference in the number of events between the 3-month periods.

Figure 4 shows that the highest number of attacks occurred in the first months of the conflict. The SHCC database shows a peak at 7 months, whereas the WHO SSA database peaks at 10 months.

Fig. 4.

Temporal trends of attacks on healthcare in occupied Palestinian territory, WHO SSA and SHCC database. 0–1 months (October and November 2023): included major events, such as Israel launched ground offensive in the Gaza strip, and the collapse of all hospitals serving northern Gaza. 7 months (May 2024): almost ceasefire. 10 months (July 2024): Announcement of new evacuation orders in Khan Younis and Rafah. 12 months (October 16, 2024): Hamas leader was killed

Fig. 5.

Temporal trends of attacks on healthcare in Lebanon, the WHO SSA and the SHCC database. 11 months (September 2024): included major events such as the assassination of one of the leaders of Hezbollah and the invasion of the grounds of southern Lebanon. 13 months (End of November 2024): the conflict reached a ceasefire

Lebanon

The highest number of attacks occurred in the 9–11 month period of both databases (WHO SSA: 63 reported attacks; SHCC 86). The lowest number of attacks was documented in the 0–2 month period of both databases (WHO SSA: 4; SHCC: 11).

Temporal trends were analyzed with the chi-square test, resulting in a p value of < 0.0001 for the WHO SSA database and < 0.0001 for the SHCC database, indicating a difference in the number of events between the 3-month periods.

The most obvious peak in the graphs was observed at 11 months. Before this peak, the number of attacks per month did not exceed 20 attacks per month in either database.

Discussion

Despite international treaties requiring the safeguarding of healthcare during armed conflicts, attacks on healthcare remain a pressing public health concern on a global scale. This study, which assessed temporal trends in attacks on healthcare across five conflict-affected countries, revealed a pattern of widespread violence against healthcare facilities and healthcare workers. Temporal trend analyses revealed significant spikes in healthcare attacks during the immediate or early phases of conflicts in four conflict-affected countries: Myanmar, oPt, Sudan and Ukraine. In Lebanon, the most significant spike was observed during a later phase of the conflict. This could likely be explained by a significant conflict escalation later in the timeline. Concerningly, it is evident that attacks on healthcare continued to occur during the entire study period in all conflict-affected countries, showing that the current fulfillment of IHL has fallen short in maintaining the special protected status of healthcare during armed conflict.

As mentioned above, all conflict-affected countries except for Lebanon experienced a significant peak in attacks during the early phase of the conflict. This is in line with the results of a previous study that assessed attacks in Ukraine in the first year of the conflict and another study that focused on attack types in oPt [15, 16]. This suggests that temporal trends in attacks on healthcare facilities in conflict-affected countries may have similar attributes. When combined with the findings of other studies that healthcare facilities are likely not mere collateral damage, this raises the question of whether (the timing of) attacks on healthcare is a deliberate strategy [7, 9].

Lebanon did not show an early peak in attacks as the conflict slumbered for some time. However, the Israeli invasion of southern Lebanon marked a major escalation of the conflict and was associated with a peak in attacks. Temporal trend analyses of other conflict-affected countries have shown that peak attack frequencies are not only associated with conflict escalations but also with other key moments, such as (attempted) cease fires.

Differences between databases

For this study, both the WHO SSA and SHCC databases were used. There were significant differences between the two databases, which can be largely explained by differences in their methodology (definitions of attacks; how are attacks reported, when and by whom; required verification levels) and reach (the WHO SSA and Insecurity Insight may have a dissimilar reach in the respective conflict-affected countries). Furthermore, the databases may be prone to political and reporting bias, as media coverage and access to information on attacks on healthcare may differ across conflict-affected countries. The most striking difference is that the SHCC includes almost 2,000 more attacks than the WHO SSA database did in the same study periods, and this difference was particularly evident with respect to Myanmar. The reasons for this are largely unknown. However, both databases are likely prone to underreporting, which has been shown in previous research. Despite the significant differences in numbers, the observed temporal trends were largely similar.

Recognizing and predicting temporal trends

Recognizing and predicting patterns of attacks on healthcare during armed conflicts is vital to prepare healthcare systems for impending or ongoing conflicts. Armed conflicts are usually preceded by rising tensions between different parties, and during such periods, significant efforts could be made to mitigate the impact of potential attacks on healthcare facilities. For example, hospitals could increase their supplies to anticipate medical material shortages, or precautionary evacuations or shelter-in-place situations could be planned and trained before disaster strikes [35].

Understanding temporal trends during armed conflicts may offer valuable insights for future preparedness strategies. By identifying when healthcare is most likely to be affected, such as during the initial phases or following major escalations of a conflict, healthcare systems and humanitarian actors can anticipate service disruptions and allocate resources more effectively [36, 37]. These findings highlight the importance of early interventions, including reinforcing and decentralizing medical supply chains, prepositioning critical assets, and initiating contingency planning for staff and infrastructure.

Furthermore, these observations underline the need for healthcare systems operating in volatile settings to not only build reactive capacity but also incorporate conflict-sensitive surveillance mechanisms. Early detection of escalating tensions, combined with temporal trend analyses of past incidents, may support the implementation of rapid-response protocols – while acknowledging the risk of false alarms [38] Such protocols could include the activation of security measures for facilities, the redistribution of critical personnel, or the temporary decentralization of services to less exposed areas. The integration of pattern recognition into health system preparedness could therefore enhance operational resilience in conflict-affected regions.

In addition, systematic documentation and comparison of temporal trends may serve a broader accountability function [39, 40]. While IHL prohibits attacks on healthcare, in practice, providing intent remains highly complex [41]. Recurrent and temporally clustered attacks raise concerns regarding potential systematic strategies [13]. Recognizing the consistent timing, recurrence, or spatial concentration of attacks, particularly during early or strategically significant phases of a conflict, provides a basis for further investigation and strengthens the argument that healthcare is not merely an incidental casualty of war. These patterns, especially when triangulated with qualitative reports, satellite imagery, or geospatial analysis, as shown in previous studies in oPt and Ukraine, can reinforce legal and policy arguments for stronger enforcement and protection mechanisms and contribute to future accountability mechanisms [41, 42].

Strengths and limitations

This study has multiple strengths. First, it is among the first studies to assess temporal trends of attacks on healthcare in multiple armed conflicts. Second, there was a significant difference in the number of reported attacks in the two databases, but the observed trends were largely similar. Therefore, the use of these two databases for temporal analyses likely makes the results more robust. Unfortunately, it was impossible to cross-link specific events from the WHO SSA database with those from the SHCC database, as the WHO SSA does not provide information on the exact locations of the attacks. Although some overlap of incident reporting is likely, a previous study comparing reported attacks in the two databases in 2017 showed that the overlap was rather minimal, suggesting significant underreporting of attacks [43]. While reporting mechanisms may have improved over recent years, underreporting is still likely a decisive factor. However, as both surveillance systems evolved (verification levels, partner networks), it could not be excluded that surveillance maturation could create artificial early spikes (media surges at conflict onset) or late spikes (improved capture).

There are several limitations to this study. As discussed before, attacks on healthcare are prone to underreporting, particularly in conflict-affected countries, as a result of repression, censorship and/or fear of retaliation. Databases that focus on attacks on healthcare will likely always underestimate the number of attacks. The most marked difference between the two databases was observed for Myanmar, with significantly more reported attacks in the SHCC databases than in the WHO SSA database, for which the exact reasons are unknown. This observation underlines the need for more comprehensive documentation and reporting of attacks, which would allow better comparison between the two databases as well as a better understanding of the implications of differing numbers for legal/advocacy use. Finally, significant peaks in the number of attacks were associated with significant conflict escalations in most of the investigated conflict-affected countries. However, further research is needed to establish causal relationships between peaks in attacks on healthcare and escalations.

Conclusions

Temporal trend analyses across five ongoing armed conflicts revealed that spikes in attacks on healthcare were consistently associated with the immediate or early phases of the conflict or with major conflict escalations. These findings underscore the need for anticipatory planning and early interventions to strengthen health system resilience during armed conflict. Although significant discrepancies exist between the WHO SSA and SHCC databases, particularly in the reported number of attacks, the overall patterns were broadly similar. This convergence suggests a degree of reliability in temporal trends and supports the broader generalizability of the findings across different data collection methodologies. Future research should further explore how the findings of this study can be used to explore the health, operational, and humanitarian consequences of attacks on healthcare infrastructure and personnel more deeply. However, to conduct more detailed research, databases with integrated geospatial data and harmonized databases are recommended.

Abbreviations

IHL

International Humanitarian Law

oPt

occupied Palestinian territory

WHO SSA

World Health Organization Surveillance System for Attacks on Health Care

SHCC

Insecurity Insight Safeguarding Health in Conflict Coalitions

IED

improvised explosive devices

Author contributions

JU and DB conceived and designed the study. JU collected and analyzed the data and performed the statistical analyses. FvO supervised the statistical analyses. JU and DB drafted the initial manuscript. All authors and nonauthor contributors have carefully reviewed the draft and given their approval for the final manuscript. JU and DB contributed equally to this work and are the joint first authors. JU, DB and FvO are the guarantors.

Funding

No funding was received for this work.

Data availability

The datasets generated and/or analyzed during the current study are available upon request. The source databases can be found online at [SSA Home | Index](.) (WHO SSA) and [Health Care in Conflict - Insecurity Insight](.) (SHCC).

Declarations

Ethics approval and consent to participate

This study involved the analysis of publicly available databases. No patients, individual participants or trials including animals were included in this study, and there was no need for METC or DEC approval. Local approval was obtained from the board of directors of VieCuri Medical Center, Venlo, the Netherlands (#2025.11).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.