Hhcsyria

Articles

Attacks against health care in Syria, 2015–16: results from a real-time reporting tool Mohamed Elamein, Hilary Bower, Camilo Valderrama, Daher Zedan, Hazem Rihawi, Khaled Almilaji, Mohammed Abdelhafeez, Nabil Tabbal, Naser Almhawish, Sophie Maes, Alaa AbouZeid

Summary

Background Collecting credible data on violence against health services, health workers, and patients in war zones is a massive challenge, but crucial to understanding the extent to which international humanitarian law is being breached. We describe a new system used mainly in areas of Syria with a substantial presence of armed opposition groups since November, 2015, to detect and verify attacks on health-care services and describe their effect.

Methods All Turkey health cluster organisations with a physical presence in Syria, either through deployed and locally employed staff, were asked to participate in the Monitoring Violence against Health Care (MVH) alert network. The Turkey hub of the health cluster, a UN-activated humanitarian health coordination body, received alerts from health cluster partners via WhatsApp and an anonymised online data-entry tool. Field staff were asked to seek further information by interviewing victims and other witnesses when possible. The MVH data team triangulated alerts to identify individual events and distributed a preliminary flash update of key information (location, type of service, modality of attack, deaths, and casualties) to partners, WHO, United Nations Office for the Coordination of Humanitarian Affairs, and donors. The team also received and entered alerts from several large non-health cluster organisations (known as external partners, who do their own information-gathering and verification processes before sharing their information). Each incident was then assessed in a stringent process of information-matching. Attacks were deemed to be verified if they were reported by a minimum of one health cluster partner and one external partner, and the majority of the key datapoints matched. Alerts that did not meet this standard were deemed to be unverified. Results were tabulated to describe attack occurrence and impact, disaggregated where possible by age, sex, and location.

Findings Between early November, 2015, and Dec 31 2016, 938 people were directly harmed in 402 incidents of violence against health care: 677 (72%) were wounded and 261 (28%) were killed. Most of the dead were adult males (68%), but the highest case fatality (39%) was seen in children aged younger than 5 years. 24% of attack victims were health workers. Around 44% of hospitals and 5% of all primary care clinics in mainly areas with a substantial presence of armed opposition groups experienced attacks. Aerial bombardment was the main form of attack. A third of healthcare services were hit more than once. Services providing trauma care were attacked more than other services.

Interpretation The data system used in this study addressed double-counting, reduced the effect of potentially biased self-reports, and produced credible data from anonymous information. The MVH tool could be feasibly deployed in many conflict areas. Reliable data are essential to show how far warring parties have strayed from international law protecting health care in conflict and to effectively harness legal mechanisms to discourage future perpetrators.

Funding None.

Published Online June 8, 2017 http://dx.doi.org/10.1016/ S0140-6736(17)31328-4 WHO Gaziantep Field Office, Gaziantep, Turkey (M Elamein PhD, C Valderrama MD, S Maes MD, Prof A AbouZeid MD); London School of Hygiene & Tropical Medicine, London, UK (H Bower MSc); Union of Medical Care and Relief Organizations,Gaziantep, Turkey (D Zedan MD); Syrian American Medical Society, Gaziantep, Turkey (H Rihawi MD); Canadian International Medical Relief Organization, Gaziantep, Turkey (K Almilaji MD); Information Management and Mine Action Programs, Gaziantep, Turkey (M Abdelhafeez MSc); Save the Children, Gaziantep, Turkey (N Tabbal MPH); Early Warning and Response Network, Assistance Coordination Unit, Gaziantep, Turkey (N Almhawish MD); and Public Health Department, Faculty of Medicine, Cairo University, Giza, Egypt (Prof A AbouZeid) Correspondence to: Prof Alaa AbouZeid, Public Health Department, Faculty of Medicine, Cairo Universiry, Giza, Cairo, Egypt abouzeida@who.int

Copyright © 2017. World Health Organization. Published by Elsevier Ltd/Inc/BV. All rights reserved.

Introduction Collecting credible data on violence against health services, health workers, and patients in war zones where access by independent observers is restricted and reporting attacks might provoke reprisals is a huge challenge.

The so-called weaponisation of health care (ie, the purposeful use of violence to restrict or deny access to health care as a strategy of war, which has been occurring with increasing intensity over the past 25 years)1 has reached unprecedented levels in Syria.2 But although reporting of attacks on health care as a specific category of violence has improved, and

some humanitarian and human rights organisations are making daring efforts to collect information on incidents,3–7 achieving strong credible data to underpin advocacy remains dogged by the absence of standardised methods.1 Collecting first-hand testimony through operations often results in limited coverage,4 whereas approaches that rely on using secondary data, such as media reports, satellite imagery, and, where possible, retro­ spective ground-truthing, are hampered by access constraints, fear for security of staff, and by reluctance of some organisations to collect or share data to a perception that this would not be consistent with their

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

1

Articles

Research in context Evidence before this study The difficulty of collecting and verifying data on violence against health is frequently described in the published work, but descriptions of tools specifically designed to do this and their methodology are few. We searched grey and peer-reviewed literature published in the past 10 years till April 6, 2017, and found substantial description of the challenges of collecting data, but little information describing the practical application of a tool. Descriptions of the impact of attacks on health care are relatively frequent and this study concurs with the general trends described.

Added value of this study The value of this Article is its detailed description of the unique methodology and the tool that facilitates further practical use and better assessment of similar contexts, and in

mandate or need for confidentiality.8,9 The absence of a systematic approach makes removing duplicate reporting and substantiating numbers problematic, while the risk of biased results is substantial if the geographical areas covered by data collection are not clearly specified. For many incidents, lack of baseline data prevents assessment of impact, whereas absence of disaggregated data (eg, by sex, age, facility, or healthcare worker type) limits interpretation.10 Even after the most rigorous and comprehensive review of data on the weaponisation of health care in Syria so far, The Lancet–American University of Beirut Commission on Syria2 found that data from only three sources contained enough detail to be combined with any confidence.

It has long been recognised that robust data are crucial to verifying attacks on health care, quantifying their effect (in terms of damage to health structure, staff, and lost services), convincing global actors of the urgent need to enforce international protections for health care, and achieving accountability for breaching them.

To address these challenges and to build on the 2010 UN General Assembly Resolution that threats to health care should be addressed,11 the 2012 World Health Assembly tasked WHO to develop a method to facilitate more reliable collection of data for attacks against health care.12 An initial tool and method was developed and, in 2015, the task of piloting this tool for field implementation was taken up by the Health Cluster in Gaziantep, Turkey, with the aim of obtaining the most credible data possible for advocacy.

The Monitoring Violence against Health Care (MVH) tool incorporates a standardised electronic data collection instrument and a verification process with anonymous real-time incident reporting. It enables a data team to receive alerts directly from a network of ground sources, use triangulation and a matching 2

the results, which support previous evidence and broaden understanding of the extent to which health care is under attack in Syria.

Implications of all the available evidence The tool described in our study offers a much needed opportunity to standardise the collection of data on violence against health care. If used widely, the tool will enable humanitarians and health-care advocates to better aggregate reports and strengthen the evidence base for advocacy. The described methods and tool will guide all interested parties to improve their systems of monitoring such attacks on health care and other similar civilian objects in conflict zones.

procedure to identify separate incidents from multiple reports, and finally assess the strength of evidence on incidents through a verification process.

In this Article, we describe the use of this tool to produce primary data for attacks on health-care facilities, ambulances, health workers, and patients across several governorates of Syria. We report findings from November, 2015, to December, 2016.

Methods

Study setting and population The Syria Response Turkey Cross-border health cluster is based in Gaziantep, Turkey, and provides a coordination mechanism for humanitarian health activities inside Syria. Its members, which include the UN and around 50 international, Syrian, and Turkish NGOs, provide cross-border assistance in ten of 14 governorates in Syria, and support 352 health facilities serving a population of approximately 5¡5 million, including 1¡9 million displaced individuals.13 The areas covered by the health cluster are mainly areas with a substantial presence of armed opposition groups, but include some locations under Syrian Government control. The MVH system was piloted in these areas from April to June, 2015, then fully implemented in November, 2015. Although network presence is higher in the northern areas in Syria with a substantial presence of armed opposition groups, there is some coverage in most governorates. Alerts from locally run facilities not supported by health cluster partners may be channelled through health cluster partners, but, in practice, areas with little or no health cluster presence contribute few reports to the system.

Data collection The MVH system consists of a network of organisations reporting incidents in real time using a standardised

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

Articles

Internal monitors

External monitors Weekly alerts report

Health cluster partners Attack flash update

Immediate alerts

First 12 h over key essential data

Verified and unverified incidents

Ad-hoc alerts report

Verification

MVH tool-alert database Compilation of alert reports for the events Monthly review of alerts

Corroboration Triangulation

MVH tool-verification database One corroborated report per incident with the verification status Data transfer Production of results, tables, and figures

Heath cluster internal monitor unit

Review committee monthly meeting

Monthly alerts report

Dissemination Monthly report, monthly infographic, and advocacy

Figure 2: Flow of data collection, data entry, and verification procedures for the Gaziantep health cluster use of the Monitoring Violence against Health Care (MVH) tool

Figure 1: Example of an initial alert post to the WhatsApp network

anonymised online reporting format linked to a curated database and a stringent verification process that ensures removal of duplicated reports and consolidation of information. All Turkey health cluster organisations with a physical staff presence in Syria were asked to participate in the MVH alert network (known as internal partners).

The network’s backbone is a 293-member WhatsApp group that allows alerts and requests for information to be shared rapidly. The first indication of an incident is often a short post to this group (figure 1). All members with physically verified information about the reported incident were then asked to complete the anonymous and confidential online alert form (appendix). Key variables include location, attack type, facility type, extent of damage, who was affected, injuries, and deaths. Physical verification means having visited the site or being present during the incident; hearsay is not valid. The alert form remains editable to the partner, so

further information can be added if available. Field staff were asked to seek further information by interviewing victims and other witnesses when possible. Within 24 h, the MVH data team triangulated the infor­ mation received and distributed a preliminary flash update of key information (location, type of service, modality of reported attack, deaths, and casualties) to partners, WHO, the UN Office for the Coordination of Humanitarian Affairs, and donors.

All aspects of the alert database were anonymised and protected with Secure Sockets Layer encryption.

Partners accessed the system with a confidential code and anonymous email address allocated centrally, but known only to the partner. The data team addressed queries via the anonymous email address; organisation names were not used at any time. Incidents were frequently reported by multiple partners. Each partner, however, could only access their own alert account and could not see the others’. If only one alert was received for an incident, the data team could contact other organisations working in the same location to see if there were other witnesses who could provide a report.

Data cleaning took place throughout the process and includes consistency checking and follow-up of missing data and questionable figures.

The team also received and entered alerts from several large non-health cluster organisations (known as external partners, who do their own informationgathering and verification processes before sharing their information).

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

See Online for appendix

3

Articles

Verified incidents

Unverified incidents

Total incidents

Number of initial alerts*

443 (2·8)

275 (1·1)

718 (1·8)

Number of sources providing information in alerts†

603 (1·4)

501 (1·8)

1104 (1·5)

Victim, eye witness, or health workers in attack location

208 (34·5%)

166 (33·1%)

374 (34%)

UN agencies, international, or national NGO

253 (42·0%)

222 (44·3%)

475 (43%)

Media or social media

85 (14·1%)

62 (12·4%)

147 (13%)

Other sources

57 (9·5%)

51 (10·2%)

108 (10%)

158 (39·3%)

244 (60·7%)

402 (100%)

Type of source for alerts (several answers possible)

Number of incidents after consolidation of alerts Availability of sources for reported incidents Reported by three or more sources

85 (53·8%)

4 (1·6%)

Verified by two sources

73 (46·2%)

26 (10·7%)

89 (22·1%) 99 (24·6%)

Verified by one source

0 (0·0%)

214 (87·7%)

214 (53·2%)

84 (53·2%)

43 (17·6%)

127 (31·6%)

6 (3·8%)

58 (23·8%)

64 (15·9%)

Primary health centre with pre-hospital trauma care

10 (6·3%)

9 (3·7%)

19 (4·7%)

Primary or community health care facility without trauma care

12 (7·6%)

18 (7·4%)

30 (7·5%)

Mobile team en route or mobile health services

26 (16·5%)

60 (24·6%)

86 (21·4%)

5 (3·2%)

5 (2·0%)

10 (2·5%)

Ambulance

14 (8·9%)

34 (13·9%)

48 (11·9%)

Unknown

1 (0·6%)

17 (7·0%)

18 (4·5%)

137 (72·5%)

153 (67·1%)

290 (69·5%)

50 (26·5%)

58 (25·4%)

108 (25·9%)

2 (1·1%)

17 (7·5%)

19 (4·6%)

Number of facilities affected, by type Hospital with trauma centre Hospital without trauma centre

Medical warehouse or health authority office

Type of attack reported (several answers possible) Aerial bombardment Non-aerial bombing or explosion Other type of attack (includes gunfire, arson, robbery, and forced closure)

Data are n (%). *Brackets represent mean number per reported incident. †Brackets represent mean number per alert.

Table 1: Distribution of reported incidents by verification status, availability and source of alerts, type of health facility affected, and type of attack

Verified incidents (n=158)

Unverified incidents* (n=244)

Patients

45 (28·5%)

23 (9·4%)

People other than patients or health workers

78 (49·4%)

72 (29·5%) 66 (27·0%)

Health workers

78 (49·4%)

Administrative staff

39 (24·7%)

22 (9·0%)

Other or unclear

25 (15·8%)

10 (4·1%)

Data are n (%).*Unverified incidents by definition feature less information on people affected, explaining the lower percentages.

Table 2: Distribution of reported incidents by category of person involved (several categories possible)

4

six key variables: date, time, geo-location, type of facility, service targeted, and health facility code.

Each unique incident then underwent a second verification process in which alerts received from internal partners were triangulated against external partner alerts. Reported attacks were deemed to be verified if they were reported by a minimum of one health cluster partner and one external partner, and the majority of the key variables matched. Alerts that did not meet these criteria were deemed to be unverified, though they are retained in the database. Nonverification is usually due to insufficient information, not evidence of false alerts.

Once unique incidents were identified, information from the different reports was pooled and assessed.

Even when a reported incident was verified, some data, such as the number of casualties, inevitably differed between reports because systematic investigations were not possible. When this occurred, the lowest number reported was retained in the final pooled report. For more qualitative data (eg, circumstances of the incident), the information retained reflected the majority opinion as long as it did not contradict information given by the partner managing the affected health service, in which case the managing partner’s information was favoured.

Once the verification process was complete, all alerts for the same reported incident were merged and incidents (verified and unverified) are entered in a separate database for analysis (figure 2).14

Analysis We described the nature of the initial alerts and unique reported incidents disaggregated by verification status and geographical location. We reported characteristics of people involved in reported incidents and outcomes of those directly harmed, by age and sex. Distribution of affected facilities by governorate was presented and, where possible, the proportion of service provision affected by reported attacks. Data were cleaned on Microsoft Excel 2016 and analysed on SPSS 21 and Stata 14.

Role of the funding source This study was part of the routine work of the Gaziantep Health Cluster; there was no funding source for this study. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Alert consolidation and verification

Results

At the end of each month, the database contained several alerts for (potentially) the same reported incidents. To obtain one report for each incident, the MVH working group (consisting of health cluster, WHO, and internal and external partners) met and identified alerts describing the same incident by matching information in the different reports against

Over 14 months (from November, 2015, to December, 2016), the MVH unit received 718 alert reports containing information from 1104 sources. The most frequent informants were organisations such as UN agencies and NGOs (43·0%), followed by victims, eye witnesses, and individual health workers reporting via NGO partners (33·9%; table 1).

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

Articles

After triangulation, 402 individual reported incidents affecting health-care delivery were identified, of which 158 (39·3%) were verified (ie, matching information was found in the alerts from at least one internal and one external partner). A further 30 incidents (7·5%) had matching information across two or three internal partner alerts, but not in an external partner report, and so were recorded as unverified. Another 214 incidents (53·2%) were reported by one internal partner only, and were also recorded as unverified (table 1).

Overall, 59·5% of verified reported incidents involved health-care facilities providing trauma care, including 84 hospitals and ten primary health-care centres with pre-hospital trauma capacity. Mobile teams and ambulances were the next most affected, accounting for 26·4% of verified incidents. Unverified incidents had a different distribution, with 38·5% involving mobile or ambulance services, 39·2% non-trauma-related facilities, and 21·3% occurring in centres offering trauma care. Aerial bombardment was reported for 69·5% of incidents (table 1).

Health workers and administrative staff were physically harmed in 74·1% of verified incidents.

Patients were physically harmed in 28·5% of verified events (tables 2, 3). Half of the verified incidents also included physical harm to non-patients attending the facility. The unverified incidents feature missing data, but the distribution is similar. The predominant source of harm in reported attacks involving health workers was the physical injury due to attack itself, but in 13 incidents, denied or delayed access to care was reported, and in three incidents, torture. There were no reports of sexual violence.

938 people were physically harmed in the incidents: 677 (72%) wounded and 261 killed (table 4). Men aged over 18 years accounted for the majority of reported deaths and injuries. However, the highest case fatality (28 [39%] of 71 children recorded, considering both verified and unverified incidents) occurred in children aged younger than 5 years. The number of people physically affected includes 225 (24%) health workers of whom 46 (20%) were killed, and 170 (18%) patients of whom 33 (19%) were killed. Deaths that occurred later due to deteriorated injury are likely to be missed. On average, 3·8 injuries or deaths were reported per verified incident compared with 1·4 per unverified incident.

Aleppo governorate experienced attacks against almost 63% of its hospitals (36 of 57), whereas all 11 hospitals in Dar’a and Quneitra were attacked at least once (table 5). More than half the hospitals in Hama, Homs, Lattakia, and Rural Damascus governorates were attacked at least once. Fewer primary health-care facilities were subjected to attack, with Aleppo most affected (19 attacks [13·1%]; table 5). Geographically, Aleppo and Idleb governorates accounted for 78·5% of verified incidents; 20% occurred in Hama, Homs,

Verified incidents (n=158) Physical violence or injury other than torture

Unverified incidents (n=244)

98 (62·0%)

115 (47·1%)

Psychological violence

9 (5·7%)

6 (2·5%)

Denial or delayed access to care or treatment

7 (4·4%)

6 (2·5%)

Torture

2 (1·3%)

1 (0·4%)

Disappearance or abduction

1 (0·6%)

2 (0·8%)

Forced to act against ethics

0 (0·0%)

1 (0·4%)

Other

7 (4·4%)

1 (0·4%)

Data are n (%). *The type of violence is only collected in relation to health workers, not other individuals affected. The total number of incidents recorded is greater than the number of incidents involving health workers because facilities usually have more than one health worker, and furthermore the latter might experience more than one instance of violence within the same incident.

Table 3: Types* of violence to which health workers were subjected (several categories possible)

Wounded

Killed

Total affected

Verified incidents Men Aged <18 years

59 (14%)

23 (15%)

82 (14%)

Aged >18 years

287 (66%)

115 (73%)

402 (68%)

Women Aged <18 years

33 (8%)

8 (5%)

41 (7%)

Aged >18 years

58 (13%)

11 (7%)

69 (12%)

437 (100%)

157 (100%)

594 (100%)

20 (5%)

24 (15%)

44 (8%)

130 (31%)

20 (13%)

150 (26%)

97 (23%)

31 (20%)

128 (22%)

Total Children aged <5 years Patients Health workers Unverified incidents Men Aged <18 years

61 (25·7%)

23 (24%)

84 (24%)

Aged >18 years

117 (49·4%)

62 (65%)

179 (52%) 44 (13%)

Women Aged <18 years

40 (17%)

4 (4%)

Aged >18 years

22 (9%)

15 (16%)

37 (11%)

240 (100%)

104 (100%)

344 (100%)

Total Children aged <5 years Patients Health workers

23 (8%)

4 (2%)

27 (5%)

7 (8%)

13 (7%)

20 (3%)

82 (8%)

15 (8%)

97 (17%)

Data are n (%).

Table 4: Distribution of victims of reported incidents, by outcome, age, and gender

Lattakia, and Rural Damascus, and less than 10% occurred in the other governorates. The latter might suggest that fewer attacks occurred, or that, particularly in mainly government-controlled areas, limited reporting occurs in the absence of MVH partner organisations (figure 3).

99 (44·0%) of all 225 hospitals and 36 (5·3%) of all 674 primary health-care facilities under MVH

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

5

Articles

Aleppo

Quneitra and Hama, Homs, Dar’a Lattakia, and Rural Damascus

Idleb

Ar-Raqqa and Damascus Deir-ez-Zor City and Tartous

Total

Hospitals In governorate

57

77

47

11

Ever attacked

36 (63·2%)

23 (29·9%)

24 (51.1%)

11 (100%)

22

11

5 (22·7%)

225

0

99 (44·0%)

Primary care facilities In governorate

145

Ever attacked

132

19 (13·1%)

69

9

7 (5·3%)

7 (10·1%)

3 (33·3%)

119

200

0

0

674 36 (5·3%)

Recurrence of attacks on the same facilities Once

32 (58·2%)

19 (63·3%)

23 (74·2%)

12 (85·7%)

3 (60·0%)

0

89 (65·9%)

2–4 times

18 (32·7%)

11 36·7%)

7 (22·6%)

2 (14·3%)

2 (40·0%)

0

40 (29·6%)

1 (3·2%)

0

0

0

6 (4·4%)

>5 times (maximum 16 times)

5 (9·1%)

0

Data are n or n (%). Total facility counts are taken from the health facilities registry file Turkey Hub Gaziantep 2015, Health mapper for Syria WHO 2013, and HeRAMS list of Primary Health Care facilities 2013. Governorates are grouped based on intensity of incidents or political similarity during the reporting period: incidents in areas with few or no health partners might be under-reported.

Table 5: Number and recurrence of verified reported attacks on hospitals and primary or mobile facilities, by governorate

Turkey

Al-Hasakeh Aleppo (212)

Ar-Raqqa (4)

Idleb (74) Lattakia (3) Hama (14)

Deir-ez-Zor (10)

Tartous (1) Mediterranean Sea

Homs (31)

Iraq Lebanon Rural Damascus (31) N Quneitra (1) Dar’a (18) Israel

As-Sweida Jordan

Incidents Verified incidents

West Bank

Figure 3: Distribution of attacks against health care by governorate and verification status

monitoring experienced verified reported attacks during the study period. Of the 135 primary and secondary care services involved, 46 (34·1%) were attacked more than once, with six experiencing between five and 16 reported attacks (table 5). Their status post-incident was known for only 247 (61·4%) of the 402 incidents. Of these, 28 6

(11·3%) of 247 incidents reportedly caused permanent closure of the affected facilities, whereas 150 (60·7%) of 247 caused the facility to be closed temporarily (defined as any facility that subsequently reopened). Only in 34 (13·8%) of 247 incidents did the facility remain fully open after the attack (table 6).

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

Articles

Aleppo

Idleb

Hama, Homs Lattakia, and Rural Damascus

Quneitra and Dar’a

Ar-Raqqa and Damascus Deir-ez-Zor City and Tartous

Total

Type of service affected Hospitals

99

33

37

15

7

0

191

Primary health-care facility

28

9

9

3

0

0

49

Mobile services

28

9

17

0

0

1

55

Other health services

57

23

16

1

7

3

107

Total reported incidents

212

74

79

19

14

4

402

Status of health-care service after each reported incident Permanently closed

17 (8·0%)

4 (5·4%)

4 (5·1%)

2 (10·5%)

1 (7·1%)

0

28 (7·0%)

Temporarily closed*

85 (2·4%)

40 (54·1%)

16 (20·3%)

7 (36·8%)

2 (14·3%)

0

150 (37·3%)

5 (5·7%)

1 (1·4%)

4 (5·1%)

1 (5·3%)

0

0

11 (2·7%)

12 (6·0%)

8 (10·8%)

3 (3·8%)

1 (5·3%)

0

0

24 (6·0%)

Displaced to other location Remained partially functional Remained fully functional

12 (5·7%)

9 (12·2%)

12 (15·2%)

1 (5·3%)

No information

81 (38·2%)

12 (16·2%)

40 (50·6%)

7 (36·8%)

0 11 (78·6%)

0 4 (100%)

34 (8·5%) 155 (38·6%)

Data are n or n (%). Areas are grouped based on intensity of incidents or political similarity during the reporting period: incidents in areas with few or no health partners may be under-reported. *Temporary closure is defined as closure of the facility for any period before reopening. Facilities where some services continued functioning are counted in the partially functioning category.

Table 6: Distribution of all reported incidents by governorate and by level of service and impact on functionality

Discussion To our knowledge, this is the first report of a standardised system producing primary data from anonymised, real-time reporting of attacks against health care by a network of operational actors. In 14 months, 718 alerts of attacks on health services were received from MVH partners. These alerts described 402 individual incidents of which 158 were verified according to MVH criteria.

Health facilities that provide trauma care experienced more reported attacks than other types of facilities. 44% of all secondary health-care facilities in the monitored locations experienced attacks. Of these, 60% were in Aleppo and Idleb, suggesting that secondary health services have become increasingly subject to attack as the war has progressed. However, the system was considerably more active in areas areas with a substantial presence of armed opposition groups, reflecting MVH internal partner presence; as such, attacks in government-held areas might not be fully captured.

Verified attacks were associated with more death and injury than unverified attacks. It is possible that verified attacks attracted greater verification resources because they were larger events.

This was the first long period of operational use of the MVH tool. Although it has not been formally assessed compared with the instrument developed and tested

by Haar and colleagues,15 actors involved consider implementation a success, but are disappointed that better data have not yet resulted in reduced attacks. The maintenance of anonymity and use of social media to speed reporting of attacks and allow investigation in real time are appreciated. The main challenge to

implementing the tool has been convincing partners to keep reporting, though they do not see any reduction in attacks following advocacy and use of their data to report to higher levels. Not having more than one partner in some remote areas meant it was not possible to verify some reported attacks, and additional challenges arose in training those reporting to use the reporting formats competently, ensuring information was complete enough to enable the verification process and the time commitment needed.

The main adaptation of the tool was to shorten the form to include only essential data and lower the threshold for verification from two internal and one external source to one internal and one external source: the former was felt to be unnecessarily restrictive.

WHO headquarters is now using the lessons learnt from our experience to revise the methods and tool for use in other emergencies.

The rigorous triangulation process of the MVH system removes the substantial problem of multiple counting and reduces the effect of potentially biased self-reports. But limitations remain, including reporting bias resulting from the political agendas of some of the reporting sources (which might result in overreporting), restricted access for verification and varying ground presence of cluster partners, and the possibility that deaths in injured people occurring after the verification process is completed are missed. On balance, these limitations are more likely to lead to under-reporting than over-reporting, since any political bias is partly taken into account by the multiple-source verification criteria. Limited health services might also mean that people with less traumatic injuries do not present and so are not recorded. Psychological injury is

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

7

Articles

40

Number of attacks [A: per month?]

35

Aleppo Idleb Hama, Homs, Lattakia, and Rural Damascus Quneitra and Dar’a Ar-Raqqa and Deir-ez-Zor Damascus City and Tartous

30 25 20 15 10 5 0 Nov 2015

Jan 2016

Feb 2016

April 2016

May 2016

June 2016

July 2016 Time

Aug 2016

Sept 2016

Oct 2016

Nov 2016

Dec 2016

Figure 4: Time trend for attacks (verified and unverified) on health care reported in Syria, November, 2015–December, 2016, disaggregated by location

not measured at all. Our measure of lost access to services can only be a crude indicator because the extent of damage to each facility is not precisely measured. The variable containing status of facilities after the incident has a high proportion of missing data (40%). However, even if every facility with unknown post-incident status were to have remained open, 53% of affected facilities would have suffered temporary or permanent loss of some or all services.

Further analysis is underway to better quantify the full effect of attacks on access to health services, and, consequently, health outcomes.

The substantial difference between verified and unverified reported incidents regarding involvement of trauma care facilities could suggest better reporting by external groups of attacks on facilities considered critical in a time of war because verification requires at least one matching external report. Analysis of time trends in incidents suggests that increased attacks on health services are associated with land operations to take over a specific location, such as the escalation in and around Aleppo city in November–December, 2016 (figure 4).

Although it is not possible to assess the level of intentional targeting, it is worrying that such a high proportion of attacks have affected the trauma services so critical for the care of injured people during war. Our analysis suggests that aerial bombardment is the main type of attack on health facilities taking place in Syria. Our system, however, does not feature specific expertise on weapons identification, and merely relies on the experience of MVH partners, who, like much of the Syrian population, have grown familiar with different types of weaponry used by combatants. Similarly, the system does not conduct forensic investigation of the verified attacks.

8

It is notable that reported attacks involved patients as victims less often than health workers or administrative staff, a finding that might in fact reflect measures to reduce patients’ exposure to attacks themselves: longterm inpatient services are often provided in separate structures called recovery houses.

The health cluster has daily experience of the efforts partners go to, to keep facilities functional, including dividing facilities between several locations so that each location hosts part of the service (eg operating theatres in one location, postoperative care in another, and delivery in a third) to try to reduce the risk that all services are affected in an attack and minimise the number of casualties. Other partners have moved underground into building basements, and in some cases, built entire new underground structures fortified to withstand aerial attacks. Partners describe moving locations sporadically, avoiding setting up any activities in a location that has been attacked before, keeping minimum staff in the facility, and avoiding presence in the health facility if there are no patients.

Statistics from Syria, to which we could directly compare the MVH findings, are difficult to find because of differences in definition and reporting periods, illustrating the continued difficulties posed by absence of standardised systems. Our findings seem to concur with reports from the Syrian American Medical Society (SAMS), which uses only first-hand infor­ mation from its ground staff and reported 172 attacks on medical facilities and medical workers from June to December, 2016, the majority caused by aerial bombardment and 42% located in Aleppo.4 During the same period, we have reported 285 incidents, 135 of which were verified. SAMS is a health cluster member

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

Articles

and it is likely that their reported attacks were also captured in the MVH database.

Our figures for the number of health workers killed during the reporting period (46) are substantially lower than the 77–104 presented by The Lancet–American University of Beirut Commission on Syria for 2016 using Physicians for Human Rights, the Syrian Network for Human Rights and the Violations Documentation Centre data. This could represent missing late deaths in our data or difference in datacapture areas. The total number of verified incidents reported by the two systems are more similar, with The Lancet–American University of Beirut Commission on Syria calculating 199 in 2016 versus 158 verified incidents in the MVH tool. Again, the discrepancy might be due to differing coverage areas.

WHO’s Attacks on Healthcare project identified 124 attacks on health-care services or health-care workers in Syria between Jan 1, and Sept 30, 2016, based on secondary sources, out of 198 attacks in 19 countries, with 366 deaths and 468 injuries overall over 9 months.16 By contrast, in 24 months from January 2012, to December, 2014, the International Committee of the Red Cross Healthcare in Danger project received reports, via direct observation and secondary sources in 11 unnamed countries, of 1586 incidents occurring within the perimeter of a health-care facility or en route to one, and at least 1050 deaths and 937 injuries.6 The difficulty of interpreting these global figures and of aligning those produced by different systems in Syria serves only to re-emphasise the need to standardise data systems and to improve methods and tools of data collection to better capture the burden of violence against health care. Adopting the MVH method and tool in other conflict zones might be helpful to do this.

Despite the efforts of the MVH project to provide a more reliable estimate of the burden of violence against health care in the Syrian conflict over the past year, and intensive advocacy on the subject by the health cluster and international entities including the UN and Security Council, no reduction in attacks has been seen.

Indeed, one of the most pressing challenges for the MVH team in Syria now is to keep partners convinced of the value of continuing to report. The most compelling argument for this is that without reliable data to show just how far some warring parties have moved away from international law protecting health care, the trend is likely to continue.

For this reason, we urge others working in conflictaffected populations to help improve available data by adopting the MVH tool with its important characteristics of anonymity, rapid alert, and stringent verification, and to redouble efforts to develop legal mechanisms through which this data can be used to bring perpetrators to account. We also urge all military forces and combatants to fully apply the fourth Geneva

Convention and its additional protocols 1 and 2,17,18 and the international community to apply pressure on combatants everywhere, in order to effectively protect health care and similar civilian services in conflict zones.

Contributors ME participated in the data collection, data analysis, interpretation, and wirting of the manuscript. HB participated in the interpretation of results and writing of the manuscript. CV, DZ, HR, KA, SM, and NA participated in the design of the methods and in data collection. MA participated in the data collection and data management for the system. NT participated in the design of the methods, data collection, and data management for the system AAZ supervised the system, participated in data management, interpretation of the results, and writing of the manuscript.

Declaration of interests The authors alone are responsible for the views expressed in this Article and they do not necessarily represent the views, decisions, or policies of the institutions with which they are affiliated. We declare no competing interests.

Acknowledgments We thank all partners in the health cluster and external non-health cluster in Gaziantep, Turkey, for their continuing efforts to reliably report incidents in addition to their arduous humanitarian work for the population. We also thank Pavel Ursu, WHO representative in Turkey, for his revision and guidance throughout the process of establishing the system and drafting this manuscript. We are grateful for the support and advice of Francesco Checchi and Bayard Roberts of the London School of Hygiene & Tropical Medicine.

References 1 Rubenstein LS, Bittle MD. Responsibility for protection of medical workers and facilities in armed conflict. Lancet. 375: 329–40.

2 Fouad FM, Sparrow A, Tarakji A, et al. Health workers and the weaponisation of health care in Syria: a preliminary inquiry for The Lancet–American University of Beirut Commission on Syria. Lancet 2017; published online March 14, 2017. http://dx.doi.org/10.1016/S0140-6736(17)30741-9.

3 Physicians for Human Rights. Anatomy of a crisis: a map of attacks on health care in Syria 2016. https://s3.amazonaws.com/ PHR_syria_map/web/index.html (accessed Feb 19, 2017).

4 Syrian American Medical Society. The failure of UN Security Council Resolution 2286 in preventing attacks on healthcare in Syria 2017. https://foundation.sams-usa.net/wp-content/ uploads/2017/01/SAMS-2286-Report.pdf (accessed May 12, 2017).

5 Human Rights Watch. Safeguarding health in conflict 2015. https://www.hrw.org/sites/default/files/related_material/ HHR%20Attacks%20on%20Hospitals%20brochure%200515%20 LOWRES.pdf (accessed Feb 19, 2017).

6 International Committee of the Red Cross. Health care in danger: violent incidents affecting the delivery of healthcare, January 2012– December 2014. https://www.icrc.org/eng/assets/ files/publications/icrc-002-4237.pdf (accessed Feb 20, 2017).

7 Medecins Sans Frontieres. Medical care under fire. http://www.msf.org/en/topics/medical-care-under-fire (accessed Feb 10, 2017).

8 Physicians for Human Rights. Anatomy of a crisis: a map of attacks on health care in Syria—methodology. https://s3. amazonaws.com/PHR_syria_map/methodology.pdf (accessed Feb 1, 2017).

9 Patel P, Gibson-Fall F, Sullivan R, Irwin R. Documenting attacks on health workers and facilities in armed conflicts.

Bull World Health Organ 2017; 95: 79–81.

10 Foghammar L, Jang S, Kyzy GA, et al. Challenges in researching violence affecting health service delivery in complex security environments. Soc Sci Med 2016; 162: 219–26.

11 UN General Assembly. Resolution 65/132. Safety and security of humanitarian personnel and protection of United Nations personnel. New York: United Nations General Assembly, 2010.

12 WHO. 65th World Health Assembly Resolutions and Decisions: WHA65.20. Geneva: World Health Organization, 2012.

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4

9

Articles

13 Humanitarian Response. Turkey Cross-Border: Health indicators reported through 4Wsâ&#x20AC;&#x201D;Q1 2016. https://www.humanitarian response.info/en/operations/stima/document/health-4ws-q1-2016 (accessed Feb 19, 2017).

14 Humanitarian Response. Turkey Cross-Border Health Cluster.

Maps and infographics. https://www.humanitarianresponse.info/ en/operations/stima/infographics (accessed Feb 19, 2017).

15 Haar RJ, Footer KH, Singh S, et al. Measurement of attacks and interferences with health care in conflict: validation of an incident reporting tool for attacks on and interferences with health care in eastern Burma. Confl Health 2014; 8: 23.

10

16 WHO. Attacks on healthcare dashboards. http://www.who.int/ emergencies/attacks-on-health-care/en/ (accessed Feb 19, 2017).

17 International Comittee of the Red Cross. Protocols additional to the Geneva Conventions of 12 August 1949. Geneva, Switzerland. https://www.icrc.org/eng/assets/files/other/icrc_002_0321.pdf (accessed April 6, 2017).

18 International Committee of the Red Cross. Convention (IV) relative to the Protection of Civilian Persons in Time of War. Geneva, 12 August 1949. https://ihl-databases.icrc.org/applic/ihl/ ihl.nsf/INTRO/380 (accessed April 6, 2017).

www.thelancet.com Published online June 8, 2017 http://dx.doi.org/10.1016/S0140-6736(17)31328-4