issue brief TLAVA TIMOR-LESTE ARMED VIOLENCE ASSESSMENT Number 4 | October 2009 Tracking violence in Timor-Leste A sample of emergency room data, 2006–08
Introduction Th e portion of recorded injuries seen at Dili National Hospital’s statistical unit and to National Hospital attributable to violence data made available by the UN Integrated Accurate data on the incidence and fell from 59 per cent in 2006 to 36 per Mission in Timor-Leste (UNMIT), which characteristics of violent injuries can be cent in 2008. Injuries caused by weapons is also conducting ongoing violent injury a powerful tool for understanding and associated with urban gang warfare— tracking. Th e Issue Brief concludes with responding to armed violence. Record arrows, knives, and machetes—dropped a review of the problems encountered keeping by hospitals, clinics, and other from 17 per cent to 6 per cent in the same in conducting this research and with a health facilities on the nature of injuries period, and injuries caused by fi sts, stones, consideration of current challenges to can provide crucial information on who and sticks dropped from 36 per cent to 24 hospital-based surveillance in Timor-Leste, is being injured, how, where, and under per cent. Th e portion of recorded injuries noting the obstacles still to be overcome to what circumstances. Data is essential for attributable to traffi c accidents rose from 41 achieve consistent, reliable, and timely data gaining a better understanding of the risk per cent to 57 per cent. collection and dissemination. factors for victimization among diff erent segments of the population and for the Th e portion of recorded injuries seen at design of interventions to reduce those risks. Baucau Hospital attributable to violence Hospital characteristics To be most useful for guiding prevention rose from 34 per cent in 2006 to 48 per and procedures activities, however, injury data should be cent in 2008, but data coding problems Th ere are six government-operated hospitals standardized, detail rich, and collected make it diffi cult to elaborate—or even in Timor-Leste with a total bed capacity systematically as part of a public health confi rm—this fi nding. Traffi c injuries of approximately 500.1 Th e Dili Hospital surveillance system. Th is in turn requires reportedly dropped from 66 per cent of all Nacional Guido Valadares (henceforth uniform information management systems, cases to 52 per cent in the same period. Dili Hospital), with 264 beds, is the best the commitment of personnel time and Data from Maliana Hospital had to be equipped. Th e fi ve regional hospitals, each resources, and motivated stakeholders excluded from the analysis due to the lack based in district headquarters—Baucau, willing to provide sustained support. of ER records for 2006 and 2007: in only Maliana, Oecussi, Suai, and Maubisse— one-third of all cases could a cause of Under an agreement with the Ministry of feature smaller facilities. All are equipped to injury be determined. Health, the Timor-Leste Armed Violence provide at least basic surgical care. Patients Assessment (TLAVA) was given access to Almost one-fi ft h of all women presenting needing urgent care are admitted through patient data from three hospitals for the at ERs in Dili and Baucau hospitals were the public patient registry or ambulance gate years 2006–08. Th e objectives were to collect recorded as victims of domestic violence; into the ER. ER staff and doctors called in and analyse the data in order to identify the proportion rose to one-third for from other departments oft en determine incidents, trends, and risk factors for armed women aged 20–39 years. initial treatment and possible transfer violence following the 2006 crisis and 2007 Men aged 15–34 years, and especially those to other units. For violent injuries, this unrest, and to assess the state of violent injury aged 20–29 years, appeared to predominate generally means surgical wards. surveillance in the country. Th is study, while in ER visits for violence, suggesting that this Hospitals track departmental caseloads and subject to limitations and caveats, presents segment of the population is at greatest risk procedures through their own statistical one of the fi rst longitudinal looks at violent of violent injury. victimization as refl ected by cases treated at systems. Th ese vary with the size and ER data is currently not suffi ciently robust Timorese hospital emergency rooms (ERs). complexity of the hospitals. Dili Hospital or systematically recorded to provide a has developed a diff erentiated statistical Aft er manually abstracting patient data reliable picture of interpersonal violence in unit with four full-time staff members. In from ERs at Dili National Hospital, Maliana Timor-Leste society. Baucau Hospital, the second-largest facility Hospital, and Baucau Hospital, and a critical Th is Issue Brief proceeds by providing an (114 beds), the medical records offi cer sees review of hospital record keeping processes, overview of the methods and objectives to statistical tasks. In Maliana Hospital (24 TLAVA fi nds that: of hospital surveillance for understanding beds), statistical and reporting duties appear Almost half of the injuries (44 per cent) violent injury trends, and describes the to be less clearly diff erentiated from other recorded at ERs in Dili and Baucau current record keeping arrangements. It clerical work. Th e variability is problematic hospitals from 2006 to 2008 were the then reviews the data collected by TLAVA for data collection and comparison, and result of violence, while 53 per cent were in three hospitals for the period 2006–08. hampered TLAVA researchers’ ability to due to traffi c accidents. TLAVA data is then compared to that of Dili draw conclusions from across hospitals.
Timor-Leste Issue Brief October 2009 | 1 Box 1 What is ‘surveillance’? Th e term ‘surveillance’, as applied in the fi eld of public health, refers to the ongoing and systematic collection, analysis, and interpretation of health information.2 Th e goal of injury surveillance—as it is with the surveillance of infectious disease—is to generate reliable data that can be used to inform rapid responses to emerging health crises (‘early warning’), and to plan long-term public policy responses. Critical to the value of surveillance is the standardized categorization and coding of medical conditions presented to health staff . Accurate, standardized coding not only allows comparisons of disease and injury (or death) seen Baucau ER setting (staged). The attending physician records patient information in a diagnostic form. at a particular hospital, but makes The ER register book, under the physician’s left arm, was the exclusive data source for TLAVA’s research. aggregation with data from other © ALDO BENINI facilities possible, facilitating the analysis of the health situation of population photocopied templates, which are sent to TLAVA researchers transferred data on groups. When collected in a timely and the statistical unit or records offi cer. Th e over 2,600 ER patients to spreadsheets. Th e ongoing way and disseminated regularly, data is aggregated into summary statistics of initial objective was to analyse injury data pooled data can be used to monitor caseloads, procedures, and—increasingly— for the entire 36-month period of January the incidence of injury characteristics, diagnoses. Th is system has limited capacities 2006 to December 2008. Th e team began such as weapons-related injuries, in the to correlate individual patient data, and the by abstracting the data for the months of population at large. Epidemiological data it generates is not normally recorded June, July, and August each year, in order to analysis can identify specifi c risk factors in a way that facilitates statistical analysis. make rapid comparisons between the events for certain types of injuries among Notably, TLAVA could fi nd no evidence that surrounding the crisis of 2006 and the unrest particular segments of the population, hospital statistics even captured the length of 2007—both of which took place in the such as knife injuries among young men summer months—and the relatively calm of patient stays. aged 15–24 years. summer period of 2008. Th e data abstraction Th e Dili Hospital statistical unit keeps its Th e tenth update of the World Health process was far more burdensome on TLAVA data in digital spreadsheets, but relational Organization’s (WHO) International and hospital staff than anticipated, however, Statistical Classifi cation of Diseases, database programs are not yet used. and plans to backfi ll the intervening months 4 known as ICD-10, is the international Th e Baucau Hospital records offi cer has had to be abandoned. standard diagnostic classifi cation started transferring handwritten report In addition, it quickly became clear that for recording health conditions in information to digital spreadsheets, but the statistical and records offi ces in the individuals who present at ERs, clinics, TLAVA has no information on the progress hospitals were either marginal (Baucau, and other acute care facilities.3 of computerization in the remaining Maliana), did not hold individual patient hospitals. Th is meant that the TLAVA data data (statistical unit in Dili), or did not do Th e extent to which patient data in Timor- collectors had to extract data from the statistics at all (records offi ce in Dili). As a Leste is digitized varies from one hospital to manual information bases—essentially the result, TLAVA data collectors fell back on another. In Dili National Hospital, computers admissions book—of the ERs. Th e practical primary data sources, i.e. ER register entries. are networked within the statistical unit, but and conceptual challenges this presented are Th e TLAVA data thus consisted initially not with other systems. Patient identifi ers discussed on pages 5–6. of 2,465 individual trauma patients seen are manually created at the registration desk in the ERs of Dili, Maliana, and Baucau or in the ER and do not always travel the full Data collected hospitals in June–August 2006, 2007, and length of an in-patient hospital stay, making 2008, aft er excluding 135 non-injury- it impossible to determine the total number To access data on injured patients treated related complaints. Some patient fi les were of patients in the hospital system or to by Timor-Leste hospitals, two TLAVA also consulted, where possible, but this was eff ectively track patient treatment. In Baucau data collectors worked at three hospitals rarely the case.5 Hospital, the individual identifi cation system during the period September 2008 As Table 1 indicates, Dili Hospital remains in its infancy. One reason for this through June 2009. Th eir access to hospital contributed the majority of the cases is that trauma care is free for most patients documentation and patient data was initially abstracted. Th is is partly due to the fact that and procedures. Without a universal patient negotiated with the Ministry of Health, the Dili had a functioning records offi ce, but billing system, hospitals have no incentive to hospital management, and, in Dili National also because of administrative challenges maintain a complete information record for Hospital, with the statistical unit. Practically, elsewhere. For example, Maliana Hospital individual patients. the data collectors carried out most of their kept few ER records from the years 2006 Some data on patients is contained in work in the ERs. To a lesser degree, they and 2007 that the TLAVA data collectors the daily reports that the departments interacted with staff of the male surgical could locate and extract. Yet in January 2008 create, with handwritten entries on ward in Dili National Hospital. some staff members were sent for training
2 | Timor-Leste Issue Brief October 2009 in patient record maintenance. Th e result is keeping, the absence of other violence cases out the set of abstracted cases, about one a crop of trauma patient records in Maliana and the very low number of people injured per cent of the recorded cases featured for summer 2008 eight times the size of with fi sts, stones, and sticks may suggest individuals unintentionally injured in non- those in previous periods. that Maliana Hospital did not systematically traffi c settings. For about two per cent of recognize low-level violence. In any event, all trauma patients, the cause of injury Table 1 ER trauma patients the Maliana dataset—18 per cent of the remained unknown or unrecorded. by hospital, June–August 2006–08 data abstracted for this research—had to Th e data collectors assigned the June–August period of year: be excluded from the substantive fi ndings classifi cation ‘domestic violence’ to six per reported below. Hospital 2006 2007 2008 Total cent of the trauma cases. Another four per Baucau 144 223 179 546 cent suff ered violence of an unspecifi ed Dili 386 523 555 1,464 Findings kind. Other violence cases were coded according to the weapon used. Th us, almost Maliana 40 46 369 455 Of the abstracted injury cases from all three a quarter (23 per cent) of all admitted hospitals, almost one-quarter (596 of 2,465) Total 570 792 1,103 2,465 trauma cases had been injured in fi ghts in were women and girls. Th e age distribution which bare fi sts, stones, or sticks were used. Th e trauma classifi cation used by the TLAVA ranged from eight months to 95 years, with Ten per cent of admissions were for wounds data collectors evolved over time, mirroring the median age being 24 years. Out-of- infl icted with arrows, knives, spears, and categories used by the Dili Hospital statistical district residents made up nine per cent machetes (grouped together in codes). unit and extrapolating from free-form of the ER trauma cases. Surprisingly, the Firearm injuries were recorded for less than diagnosis and cause-of-injury text elements. better-equipped Dili National Hospital saw one per cent of the trauma patients, but Early on, an irreversible choice was made fewer out-of-district trauma victims (10 per some hospital staff have indicated that some of mixing cause of incident and weapons cent) than Baucau Hospital (12 per cent). fi rearm victims stay away from hospitals, categories. Aft er reclassifying some categories Th is may refl ect the fact that Baucau is the resulting in under-reporting.6 and collapsing rare ones into related only hospital for the entire eastern region categories, the TLAVA team settled on the of Timor-Leste; stronger vehicular traffi c Trends in violence and weapon use categorization shown in Table 2. growth and more intense political and gang Th e composition of the ER trauma patients It must be noted immediately that the violence in Dili may have also played a part. changed signifi cantly over the three mixing of incidents types and weapon Th e Maliana Hospital ER did not tend to observation periods (see Tables 3 and 4). types in a one-dimensional category set is receive any trauma patients from outside In Dili National Hospital, the fraction of analytically inappropriate. For example, Bobonaro District. One explanation for violence victims fell steadily, from 59 per a case of domestic violence in which a this may be that it is situated not far from cent in 2006 to 36 per cent in 2008. Th is knife was used cannot be placed in both Maubisse Hospital, which was not included is logical, given the violent political events categories. In such cases, data collectors in this study. of 2006, aft er which the political situation broadly stabilized—despite the March 2007 needed to choose one or the other—and Injuries in either case, valuable information is lost. attacks on the president and prime minister. Furthermore, categories varied greatly from Th e following fi ndings are drawn exclusively In parallel, Dili saw a rapid increase in the hospital to hospital. from Dili and Baucau, as Maliana Hospital number of motor vehicles and, with them, data had to be excluded for the reasons an increase in traffi c accidents. Maliana Hospital presented special, noted above. Th e majority of trauma insurmountable problems: the data patients recorded were victims of traffi c Table 3 Violence as a percentage of collectors found the ER records to be so accidents (53 per cent). Almost half (44 all trauma cases sparse that they could make a determination per cent) of the 2,010 cases presenting for of the incident type in only one-third of all trauma were as a result of violent incidents. Hospital cases. Here, no domestic violence case was Th e proportion of violence victims was June–August Baucau Dili Combined recorded as such. While the data may not only slightly higher in Dili (45 per cent) 2006 34.0% 59.1% 52.3% be representative because of poor record than in Baucau (42 per cent). Rounding 2007 41.7% 43.8% 43.1% 2008 48.0% 36.2% 39.1% Total 41.8% 44.9% 44.1% Table 2 ER trauma patients by incident type and hospital (Baucau, Dili, Maliana), aggregate 2006–08 Th ese developments cannot explain the Hospital evolution of the Baucau Hospital cases, Type of incident Baucau Dili Maliana Totals however, where violence increased from N% N% N% N%about one-third to almost one-half the ER Traffi c accident 318 58 743 51 113 25 1,174 48 trauma load from 2006 to 2008. While the Baucau ER coordinator could note only Other accidents 0 0 24 2 10 2 34 1 that domestic violence cases—particularly 5 6 0 5 Domestic violence 30 95 0 125 wives injured with machetes—remained Other violence 65 12 18 1 0 0 83 3 frequent, the data points to a worsening Fists/stones/sticks 64 12 400 27 3 <1 467 19 trend for violence generally. Teasing out the Arrows/knives/machetes 58 11 140 10 25 5 223 9 nature of the increase is challenging because Firearms 11 2 5 <1 0 0 16 <1 of the confl ation of injury- and weapons- Unknown 0 0 39 3 304 67 343 14 type coding. While the proportion of female victims coded in either the ‘domestic Total 546 100 1,464 100 455 100 2,465 100 violence’ or ‘arrows/knives/machetes’
Timor-Leste Issue Brief October 2009 | 3 Table 4 ER trauma patients by incident type and hospital, 2006–08 of men (211 females vs 675 males). For (Baucau and Dili) accident-related trauma patients (most due to motor vehicle accidents) and those of
BAUCAU DILI unknown cause, females aged15–24 years TYPE OF INCIDENT 2006 2007 2008 2006 2007 2008 TOTAL Traffi c accident Cases 95 130 93 158 270 315 1,061 represent the most elevated risk. %66.0 58.3 52.0 40.9 51.6 56.8 52.8 Almost one-fi ft h (19 per cent) of all women Other accidents Cases 000 061824 visiting ERs in Dili and Baucau were %0.0 0.0 0.0 0.0 1.2 3.2 1.2 Domestic violence Cases 81210214331125coded as domestic violence victims (87 %5.6 5.4 5.6 5.4 8.2 5.6 6.2 of 453 cases). Th e highest risk age groups Other violence Cases 17 24 24 1 10 7 83 were 20–24 and 25–39 years, in which %11.8 10.8 13.4 0.3 1.9 1.3 4.1 the percentages of injuries attributable to Fists/stones/sticks Cases 4 32 28 137 131 132 464 domestic violence were 29 per cent and 32 %2.8 14.4 15.6 35.5 25.1 23.8 23.1 per cent, respectively. Th e number of cases Arrows/knives/ machetes Cases 19 21 18 67 42 31 198 %13.2 9.4 10.1 17.4 8.0 5.6 9.9 of domestic violence increases in 2007, Firearms Cases 146 230 16slightly in Baucau and more dramatically in %0.7 1.8 3.4 0.5 0.6 0.0 0.8 Dili, before dropping again in 2008. Unknown Cases 0 0 0 0 18 21 39 % 0.0 0.0 0.0 0.0 3.4 3.8 1.9 Comparisons with other Total Cases 144 223 179 386 523 555 2,010 %100.0 100.0 100.0 100.0 100.0 100.0 100.0 data sources Dili Hospital statistical unit categories increased slightly from 2006 Age as a risk factor for men and women to 2008, much more signifi cant was the In 2006 and 2007 the statistical unit of Dili Given problems of data coding, identifying sevenfold increase in cases of lightly armed Hospital produced, as part of its monthly risk factors for the population is challenging. fi ghting (‘fi sts/stones/sticks’). Th ere is at reports, tables of accident- and violence- One exception is for specifi c age groups least one observation that may suggest that related trauma patients. Th e counts were for men and women. Figure 2 presents the 2006 data for this category for Baucau disaggregated according to gender and violent injuries for males by fi ve-year age Hospital is an anomaly: the portion of cause (see Figure 3). Th e data for the second patients seen for arrow, knife, and machete cohort. It shows that from age 15 to 39 half of 2007 was unfortunately lost due injuries in 2006 was about the same in both years, males are highly over-represented in to a computer virus attack at the hospital Baucau (13 per cent) and Dili (17 per cent). trauma hospitals—especially in the 20–24 statistical unit. Since 2008 some violence Th us, the causes of a trend in Baucau—if it age group—compared to other age groups statistics have been appended in a less is real—remain unexplained. within the population. Similarly, males aged systematic way in monthly spreadsheets of 20–34 years are at increased risk of accident- patient fl ow data. Trends in weapon use are more readily related trauma (not shown here). Population While informative, the data presents discernable in the data from Dili Hospital data comes from the 2004 national census.7 (see Figure 1). Lightly armed violence cases a number of questions. Th e violence remained the major category and remained Th e age distribution of female violence associated with the 2006 crisis is almost steady (137 in 2006; 132 two years victims examined is somewhat fl atter than documented to have taken place between later). Th e use of arrows, knives, and that for males, but with elevated risk in the the end of April and the end of June. But machetes—the stock-in-trade weapons of 20–29 years age group. Th is group includes according to Dili Hospital data, aft er an urban gangs—dropped considerably, from newly married women—potential primary initial surge in trauma load in May, the 67 to 31 patients. No fi rearms victims were victims of domestic violence—and young number of trauma cases—both violence and recorded in Dili Hospital in 2006. Gunshot women who were possible victims of sexual traffi c related—dropped by 50 per cent. wound victims may have died or avoided violence. Th e absolute number of female Th e chaos and fear surrounding the crisis hospitals and received treatment elsewhere violence victims is about one-third that may partly explain this. Many victims did not give their real names, or gave diff erent names at diff erent points of treatment, and Figure 1 Intentional injuries vs traffi c accident injuries, Dili Hospital, 2006–08 sometimes self-discharged in the middle of the night. Anecdotally, many violence 125 victims claimed to have been accident Traffic accident victims. Furthermore, access to the hospital injuries 100 may have been dangerous until autumn, and Fists/stones/ many of the injured are said to have sought 75 sticks emergency treatment in local clinics that did not require them to cross lines of hostilities. Arrows/knives/ In 2006 the hospital staff learned not to ask 50 machetes too many questions of injured patients and their anxious entourages. As a result, data Firearms 25 on patients’ residence is less complete for the
Patients seen in Dili Hospital ER Patients Dili records than elsewhere. 0 JUNE JULY AUG JUNE JULY AUG JUNE JULY AUG Th e erratic oscillations in the gender ratio 2006 2007 2008 of victims are also suspicious. If accurate, the low representation of male victims from
4 | Timor-Leste Issue Brief October 2009 Figure 2 Age distribution of male violence victims, 2006–08, and illustrate some of the diffi culties of hospital- 2004 male population distributions based surveillance. Th ey also limit the usefulness of comparisons with the TLAVA 18% data. In retrospect, the decision to select the Male violence victims in Baucau and Dili ERs 16% period June through August in each year is Male population, 2004 14% less than ideal, given the known dynamics of 12% the 2006 and 2007 political events. For 2006, 10% where a comparison is possible, TLAVA 8% violence counts closely matched the hospital fi gures. For accidents, TLAVA extracted 6% signifi cantly fewer cases (see Figure 4). 4%
% in particular age group 2% UNMIT 0% In response to the crisis of 2006, UNMIT
85+ launched its own national violent incident 00–04 05–09 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 monitoring system in late 2007, based on Age groups incidents recorded by police and armed forces. UNMIT staff considered the weekly Figure 3 Trauma patient data compiled by Dili Hospital, incident tallies to be somewhat complete January 2006–June 2007 from March 2008 onward.8 Although TLAVA was unable to access the basic 250 6 incident records, UNMIT gave a public presentation detailing monthly incident totals for the period January 2007 through 200 5 October 2008, which TLAVA obtained.
150 4 UNMIT’s reported trends do not match TLAVA’s counts of violence victims seen Ratio in the Dili and Baucau hospitals during 100 3 Trauma cases 2007 and 2008. In 2007 a wave of arson attacks, following elections in July and the 50 2 announcement of a new government on 6 August, sent the UNMIT count soaring. 0 1 Assaults increased slightly. Although civil JUL SEP FEB FEB JAN JAN JUN JUN APR APR DEC OCT AUG NOV unrest engulfed both Dili and Baucau, ER MAR MAR MAY MAY patient counts from these hospitals do 2006 2007 not refl ect this. During 2008 the increase Accidents Violence Male-to-female ratio, all trauma patients in UNMIT-reported incidents was much slower. Th e TLAVA count of violence Figure 4 Comparison between Dili National Hospital statistics and TLAVA victims treated in the hospitals only notes a data, June–August 2006 small increase for August 2008. As was true of the Dili National Hospital 100 statistical unit and of the TLAVA data collection, UNMIT too made changes to its category set during the period of shared 80 data. For example, domestic assault became Hospital statistics a distinct category in October 2007. It Cases was used in the last quarter, but then was 60 suspended between January and April 2008. Disorderly behaviour had been included as TLAVA an incident type since the system was created, 40 count JUNE JULY AUGUST JUNE JULY AUGUST but instances were not eff ectively counted until May 2008. Th e fl uid nature of category Accidents Violence systems among TLAVA, UNMIT, and the hospitals alike complicated surveillance. April through July 2006 could be because that month. Th e peak in reported accident Challenges encountered they feared being intercepted by hostile victims in February 2007 may be the result groups en route to the hospital, followed by of incorrect accounting of January or March Th e data collected and presented here, while a pent-up demand for trauma services once cases. Again, the gender ratio changes subject to limitations and caveats, presents tension in Dili abated. However, starting in dramatically during the six months for one of the fi rst longitudinal looks at violent December 2006, the pattern of anomalies which data is available in 2007. victimization in Timorese ERs. It represents appears to be more closely related to data a sustained process of cooperation between management issues. Th e statistical unit Th e hospital’s eff orts to put a statistical government and civil society, with resources did not report a single violence victim for stamp on the fast-changing realities and time committed by both the public and
Timor-Leste Issue Brief October 2009 | 5 private sectors. Th e project demonstrated Overview of hospital hardware donations, a dialogue of which the that quick mobilization, fl exible tasking, hospital support unit in the ministry was not and reasonably priced manpower can data management in aware. Th e information specialists in this enable partnerships to defi ne and pursue Timor-Leste unit approached information management important research goals that otherwise from a data quality angle and from the Th e cornerstone of injury surveillance is would be unlikely to fi nd sponsors. But it necessity of supporting the statistical staff individual record keeping. Unfortunately, also struggled with the challenges of staff and other information producers with a none of the three hospitals examined capacities, language, cultural divisions, and coherent skills development programme.9 maintains individual-level data. Th is is lost other obstacles. In this evolving institutional environment, in the organizational memory, except for it was not always clear who was responsible One challenge became clear at the outset, painstaking reconstruction from hardcopy for surveillance-related activities. Th e which was that data was not available records kept, if at all, outside the statistical statistical and ER staff expressed no explicit in a form that could be readily utilized unit. Th e major product of the Dili National surveillance concept. in the project. As a result, TLAVA staff Hospital statistical unit (and of the Baucau was made responsible for coding ER records offi cer’s statistical work) is a Coordination and training challenges cases. Because hospital records were monthly report addressed to the hospital abound. For example, in spring 2009 the unexpectedly incomplete or presented director and the Ministry of Health. Th e statistical unit in Dili and the records confl icting data, the data collectors were report comprises some fi ft y pages of tables, offi cer in Baucau started coding diagnoses forced to make decisions for which their each the result of a diff erent spreadsheet in according to WHO’s ICD-10 (see Box 1).10 experience and training did not prepare which the unit manually compiles data from A signifi cant portion of the cases could not them. TLAVA staff and hospital statistical departmental sources. be coded using the coding sheets provided workers also changed trauma categories by the ministry, apparently because the Th e management value of the monthly on an ad hoc basis, compromising the options were too limited or in part were not reports is not immediately clear; in one utility of the dataset. Some categories, understood. Th e coders had no personal hospital, staff volunteered that since this such as traffi c accidents, were gradually contact with the authors of the codes or with reporting was resumed aft er Timor’s or suddenly refi ned, while violence was the consumers of their statistics. Eventually, independence, they had never received any left undiff erentiated. Loss of data and the director general and the WHO feedback on it. In fact, as far as TLAVA was formatting changes in the statistical units representative put hospital information able to determine, little comparative use is frustrated the ability to cross-validate management reform on hold, pending the made of the statistics. We could not discover TLAVA results with hospital reports. arrival of a WHO specialist.11 whether there was more to an annual report TLAVA and hospital statistical workers than the collection, in one Excel workbook, It should be noted that the Ministry of also had to overcome major language of monthly and quarterly reports. Health does practice active surveillance issues. Both sides were made to use for infectious diseases. Th e line runs Th us, knowledge resides primarily with templates in English and Portuguese, not almost completely outside the hospital individual staff rather than in transferable the fi rst languages of either the hospital system, from polyclinics through district systems. For example, the records offi cer in or TLAVA staff , while daily ward reports health coordination offi ces to the Centre Baucau Hospital is locally recognized for and register entries were written in an for Communicable Diseases in the his long institutional memory. In Dili, the even greater variety of languages. Th e ministry. In May 2009 TLAVA learned head of the statistical unit was transferred data collectors’ command of English was that districts reported daily on new cases out of the hospital, and with him left the weak and translation slowed down the of H1N1, Dengue fever, and the human understanding of some of the codes that he collection process. Th e struggle to translate immunodefi ciency virus (HIV). Th e Baucau had developed. As a result, his successor probably compromised the reliability of Hospital ER coordinator made similar could not make sense of the data from that some data to an unknown degree. In fact, reports if and when such cases presented at period. Overall, the general appreciation case determinations, amid conceptual the hospital.12 Horizontal coordination with for data collection systems and their utility and language ambiguities, absorbed a the district offi ce, in the case of the Baucau appeared modest among hospital staff . considerable part of the time and energy of ER coordinator at least, was chiefl y about TLAVA workers—and apparently also of Th is is not for lack of interest. At all Dengue fever cases, which would prompt their hospital counterparts—for some time levels, from the statistical workers to the residential spraying. before any statistics were produced. director general of the ministry, there is Interpersonal violence is not subject to the Compounding this was a lack of systematic an awareness that the current hospital same surveillance systems in Timor-Leste. professional oversight and supervision patient information systems, while basically While domestic violence may be a growing of the staff , and of integration with other functional, can and should be improved. cause of concern within some government information management projects. For However, with a plethora of recovery and agencies, this has not yet aff ected observable example, TLAVA staff had diffi culty getting development projects straining coordination data collecting and reporting activities. In accepted in Baucau and Maliana hospitals, in the health sector, a number of poorly fact, the ability of researchers and offi cials and they were not able to benefi t from connected initiatives militate against a to review domestic violence incidents has improvements in coding and spreadsheet unifi ed approach. Th e statistical workers been blocked by recent changes to the design being implemented in hospital were anxious primarily to improve their monthly report template, which between statistics. Th e latter innovations were personal computer skills and to support 2007 and 2009 shift ed from recording the geared towards the production of timely colleagues in other units recently equipped causes of trauma to recording departmental and accurate reports on patient fl ows, with computers (e.g. the pharmacy), but caseloads. procedures, and diagnoses. Neither near- felt that they were too marginal to be heard real-time surveillance nor violence was a on the larger questions of systems design Linguistic and professional fractures central concern in the hospitals’ routine and development. Dili National Hospital that run through the healthcare system work. was talking with a foreign embassy about complicate matters further. Document
6 | Timor-Leste Issue Brief October 2009 templates are mostly in Bahasa Indonesian of intense violence, the trauma caseload the hospitals, TLAVA’s concern with armed or Portuguese. Entries come in several borne by the hospitals may actually fall. violence is unlikely to supply a unique focus. more languages, including, in descending Th is happens when the people caring for Other related social problems may generate order of prominence, Bahasa, Portuguese, the injured decide against taking them to stronger and more continuous collaboration English, Tetun, and Spanish. In addition, hospitals, or when service capacity plummets. with the hospitals, as well as more coherent international doctors and administrators attention to the data produced there. A Moreover, trauma statistics are a minor working in the Timor-Leste healthcare case may be made to have the hospitals by-product of hospital information system use their own languages participate not in one, but in several management in Timor-Leste. Capacity conversationally, including Chinese, thematically distinct surveillance networks. development eff orts are being pursued Filipino, and Spanish. During the original Hospitals are already connected, although by a number of groups scattered across admission of a patient to hospital, the cause apparently on the margins, to contagious hospitals, the Ministry of Health, and the of injury may be stated in the ER register, disease surveillance. Another candidate WHO country offi ce, but these groups are but this information is not carried through focus is domestic and sexual violence, weakly coordinated among themselves. to reports made by ward physicians (who which has a strong network of concerned TLAVA, during its engagement, was able may focus on, for example, a deep cut organizations, including, but not limited to build signifi cant rapport only with one wound, but not the domestic violence that to, hospitals.16 While each prominent focus of the less infl uential groups, the statistical may have caused the injury). Th is makes may benefi t from one driving organization unit in Dili Hospital. Nothing indicated that coordinating its kindred network partners, it impossible for the statistical unit to code eff orts at stronger information management the technical side of surveillance will require cases. It is fair to say that, even without had a surveillance focus, let alone one on patient and humble work on the basics. consistent patient IDs, departmental violence. Th e public health surveillance Th e natural leader for this is the Ministry caseload statistics are presumably reliable. lines that the ministry has set up are of Health, assisted by the partners that it But nobody can know how many diff erent largely running outside the hospitals. In considers appropriate for the task. individuals are cared for by the hospital in addition, community-based surveillance 13 total. Th is ignorance extends to trauma of confl ict and violence is being promoted In the fi nal analysis, TLAVA partnered with cases, and thus to our research. in a partnership between the government organizations that provide vital services to and the national NGO Belun.14 Th is victims of armed violence, but do not yet fi ll Refl ections arrangement is supposed to respond to risks a prominent role in violence surveillance. Moreover, selective admissions during Injury surveillance is a public health tool of armed violence much more rapidly than peak periods of political unrest suggest that with great potential for informing violence hospitals are equipped to do. trauma caseloads, viewed in isolation, are prevention activities. If properly recorded, Timor-Leste earned plaudits for the not suffi cient instruments of surveillance. collated, and analysed, surveillance data can speedy and eff ective way in which health help identify risk factors for specifi c groups system leadership was transferred, soon Yet violence research in hospitals has and trends over time, and even monitor the aft er independence, from international value for eff ective health information health impacts of public policy initiatives. In NGOs to the government. Improved management. If hospital statistical workers countries where public health infrastructure health information systems have been can be helped to piece together the is weak or overburdened, however, the on the agenda of the Ministry of Health trajectories of trauma patients through ability of health professionals to record and continuously since then.15 Despite this the heath system, their data management make use of injury surveillance may be progress, the needs in this area are still controls should be able to handle individual challenging. vast. Whatever the intent and shape of patient-based data for the entire hospital population. As TLAVA came to understand TLAVA’s analysis of trauma patients appears information systems created or reformed during nearly nine months of working in to confi rm a common perception that in the hospitals, there is a need to reinforce hospitals, a more sophisticated system will violence in Dili decreased between 2006 and elementary data management skills, as well as design and implement the kinds require progress on two fronts: skills and 2008. Th e same research, however, suggests systems to integrate data across hospital that violence in Baucau increased over the of formats, tools, and procedures that the available workforce will understand. units, and a viable coalition of concerned same period. Because of problems related partners wanting to use this information in to data recording and collection, it is not Moving public health surveillance for addressing important societal issues. known if this is refl ective of actual violence. trauma forward in Timor-Leste will require Meanwhile, data from Maliana Hospital investment in common conceptual and Notes was insuffi ciently recorded for substantive operational foundations. At a minimum, the results to be drawn. Th is, at most, is a hint patient identifi ers need to be standardized Th is Issue Brief was produced by Aldo Benini, who has worked for the International Committee of that the smaller provincial hospitals are in across contexts—e.g. from police reports the Red Cross and the Global Landmine Survey. the very early stages of building up their to ER records to court documents—as a He has a Ph.D. in sociology from the University patient information capacity. precondition for relating attributes from of Bielefeld, Germany, based on fi eld research on community development in West Africa. one context to those of others. Equally, In fact, in none of the three hospitals basic data entry and soft ware skills are 1 A study of fi ve out of the six hospitals calculated examined is public health surveillance the total beds at 450 in 2007 (Ministry of needed at several nodes of collaboration if properly understood. Th e process was too Health and Netherlands Royal Tropical the information produced is to be valid and slow, with these results coming out almost Institute, 2008, p. iv). Data for the sixth hospital reliable. (Maubisse) was not reported. one year aft er the end of the last observation 2 WHO (2001), p. 11. Th e full WHO defi nition period (August 2008). Even if it had been Surveillance must be driven by a strong reads: ‘Surveillance is the ongoing, systematic available almost instantly, hospital-based stakeholder coalition, and making the collection, analysis and interpretation of health trauma data would not meet the early surveillance system sustainable will require data essential to the planning, implementation, and evaluation of health practice, closely warning function expected of public health attracting and engaging these stakeholders. integrated with the timely dissemination of surveillance. As we have seen, in periods Given how few fi rearm victims are treated in these data to those who need to know. Th e
Timor-Leste Issue Brief October 2009 | 7 fi nal link of the surveillance chain is in the 16 Th e Offi ce of the Secretary of State for the Suffl a, S., A. van Niekerk, and N. Duncan, eds. application of these data to prevention and Promotion of Equality, previously known 2004. Crime, Violence and Injury Prevention control. A surveillance system includes a as the Offi ce for the Promotion of Equality, in South Africa: Development and Challenges. functional capacity for data collection, analysis takes a strong interest in domestic violence. It Tygerberg: Medical Research Council, and dissemination linked to public health promotes the expansion of hospital-based safe University of South Africa. programs.’ spaces for victims, not only in Dili, but also Timor-Leste. 2006. Timor-Leste Census of 3 WHO (2003). in the regions. With a seat on the Council of Population and Housing: National Priority Ministers, this body wields considerable power 4 A planned violent injury costing exercise also Tables. Dili: National Directorate of Statistics and could well spearhead domestic violence- had to be abandoned when it became clear and UNFPA. that it was not feasible to correlate surgical focused surveillance (AusAID, 2008, p. 25). ward patient data with ER records. Th e labour- Since 2002, one such shelter facility has been United States Bureau of Justice Statistics. 2009. intensive work succeeded in correlating only 41 operated within Dili Hospital by the NGO ‘Crime and Victims Statistics.’ Washington, cases. PRADET. DC: US Department of Justice, Offi ce of Justice 5 Locating and in some cases reviewing patient Programs, Bureau of Justice Statistics. Accessed fi les required the cooperation of busy hospital Bibliography 7 July 2009.
TLAVA project summary TLAVVA Th e Timor-Leste Armed Violence Assessment (TLAVA) is an independent research project overseen by ActionAid Australia (formerly Austcare) and the Small Arms Survey. Designed in consultation with Th e project is supported by Australian Agency for International public and non-governmental partners, the project seeks to identify and disseminate concrete entry Development (AusAID). points to prevent and reduce real and perceived armed violence in Timor-Leste. Th e project functions as a Dili-based repository of international and domestic data on violence trends. From 2008 to 2010, the Credits TLAVA is to serve as a clearinghouse for information and analysis with specifi c focus on: Design: Go Media Design the risk factors, impacts, and socioeconomic costs of armed violence in relation to population Editorial support: Emile LeBrun, Robert Muggah, Celia Paoloni, health—particularly women, children and male youth, and internally displaced people; and Lyn Wan the dynamics of armed violence associated with ‘high-risk’ groups such as gangs, specifi c communities in aff ected districts, petitioners, veterans, state institutions, and potential triggers such Contact as elections; and For more information, visit www.timor-leste-violence.org the availability and misuse of arms (e.g. bladed, home-made, or ‘craft ’ manufactured) as a factor or contact contributing to armed violence and routine insecurity. [email protected] Th e project’s objective is to provide valid evidence-based policy options to reduce armed violence for the Timorese government, civil society, and their partners. Th e project draws on a combination of methods— from public health surveillance to focus group and interview-based research—to identify appropriate formerly priorities and practical strategies. Findings are released in Tetum and English. TLAVA Issue Briefs provide timely reports on important aspects of armed violence in Timor-Leste, including the availability and distribution of small arms and craft weapons and election-related violence.
8 | Timor-Leste Issue Brief October 2009