Timeliness of epidemiological outbreak investigations in peer-reviewed European publications , January 2003 to August 2013

Timely outbreak investigations are central in containing communicable disease outbreaks; despite this, no guidance currently exists on expectations of timeliness for investigations. A literature review was conducted to assess the length of epidemiological outbreak investigations in Europe in peer-reviewed publications. We determined time intervals between outbreak declaration to hypothesis generation, and hypothesis generation to availability of results from an analytical study. Outbreaks were classified into two groups: those with a public health impact across regions within a country and requiring national coordination (level 3) and those with a severe or catastrophic impact requiring direction at national level (levels 4 and 5). Investigations in Europe published between 2003 and 2013 were reviewed. We identified 86 papers for review: 63 level 3 and 23 level 4 and 5 investigations. Time intervals were ascertained from 55 papers. The median period for completion of an analytical study was 15 days (range: 4–32) for levels 4 and 5 and 31 days (range: 9–213) for level 3 investigations. Key factors influencing the speed of completing analytical studies were outbreak level, severity of infection and study design. Our findings suggest that guidance for completing analytical studies could usefully be provided, with different time intervals according to outbreak severity.


Introduction
The International Health Regulations (2005) stipulate that each State Party is required to ensure they have the capacity to respond 'promptly and effectively to public health risks', such as outbreaks of communicable diseases [1].The timeliness of outbreak investigations is vital for containing outbreaks and preventing further cases, minimising the impact of an outbreak on both patients and health services, yet despite this, no guidance or standards currently exist regarding what might be considered a timely investigation.
A national Field Epidemiology Service (FES) was formed within Public Health England (PHE) in April 2013, following reforms to the health system in England under the Health and Social Care Act 2012 [2].The FES aims to improve the consistency of high-quality epidemiological investigations.Hence its formation prompted the consideration of whether specific guidance on the timeliness of outbreak investigations was feasible and, if so, what should be recommended.
Current guidance from PHE [3] states that outbreak reports should be completed within 12 weeks of the formal closure of an outbreak, a common standard for outbreaks within England classified as level 2 or above (Table 1).Such outbreak reports are compiled for internal purposes, to detail the steps within and results of an outbreak investigation, they may or may not include an analytical study (case-control or cohort, for example) and will not necessarily lead to publication in a peer-reviewed journal.No further guidance or standards currently exist to inform what might be considered high quality in terms of timeliness.We sought to review specific time intervals within epidemiological investigations, those from declaration of an outbreak to hypothesis generation and hypothesis generation to the availability of analytical results to inform the actions of relevant authorities.This review aims to summarise these time intervals in published outbreak investigations with cross-regional, national or international implications (using PHE definitions for level 3, 4 and 5 outbreaks, as in Table 1) to assess the feasibility of developing PHE guidance for the timeliness of analytical studies in outbreak investigations and whether separate standards for different outbreak levels would be appropriate.Such guidance or standards could be used to inform service improvement and/ or monitor performance in England and could be similarly developed in other countries.

Methods
MEDLINE, Embase and Eurosurveillance were searched using the search terms 'outbreak', 'case-control' and 'cohort' to identify outbreak investigations in Europe published between 2003 and August 2013 and which included an analytical epidemiological study.Papers were included if they met the following criteria: they reported on an outbreak occurring within Europe; were published since 2003 and the outbreak occurred in 2000 or later; were deemed to be level 3 or above (Table 1); were available in English, French, Spanish, German, Greek or Italian; and an analytical study was carried out (e.g. a case-control or cohort study) (Box).
Time intervals from outbreak declaration to hypothesis generation and from hypothesis generation to availability of analytical results were ascertained from peerreviewed publications retrieved by our search.If the intervals were not explicitly stated, estimations were made, where possible, based on the dates reported.The date of commencing an analytical study was assumed to be within one day of hypothesis generation (given an analytical study cannot commence without a hypothesis having been defined).Analytical results were assumed to be available one day following the end of the analytical study period, i.e. the period over which data collection was reported to have occurred and following which results would be available to inform action and control measures.PHE outbreak level definitions were applied to all reviewed outbreaks: judgement of the outbreak level was based on the geographical spread of cases, the involvement of national agencies in the investigation and the potential severity of population impact.In England, PHE local centres are responsible for establishing outbreak control teams and leading investigations that affect their local population, with support from the FES.Investigation into nationwide outbreaks or outbreaks with wider impact or greater severity are led nationally by the National Centre of Disease Surveillance and Control or the FES, working as part of a team with local PHE centres and other agencies as relevant.
Outbreaks of gastroenteritis infections were classified into mild to moderate or severe according to the causative organism.Severe infections were those with a recognised risk of serious long-term complications or death.Hypothesis generation and analytical study time intervals were compared by outbreak level, study design, type of infection and number of cases.The median number of days was calculated for each time interval considered.Pearson's correlation coefficient was calculated for the association between the number of cases identified and time intervals within the investigation.
The search and analysis of papers was carried out by one researcher (EV) with regular communications with both co-authors to discuss assumptions and categorisation of studies.Where more than one study related to the same outbreak, the investigation that gave a more complete overview (i.e. more cases) was selected for inclusion.

Results
The search yielded 1,522 publications, which were reduced to 1,208 following removal of duplicates.After a review of the abstracts, 290 full-text papers were selected for further screening.Application of the inclusion and exclusion criteria led to 86 studies being selected for this review .Of the selected results, 63 were classed as level 3 outbreaks, 22 level 4 and one level 5. Given the small number of level 5 outbreaks, these were combined with level 4 outbreaks for the analysis.

Distribution by country
The countries with the highest number of outbreaks, with peer-reviewed reports meeting the inclusion criteria were the United Kingdom (UK), Germany and the Netherlands ( Germany was the location of the level 5 outbreak investigation that was reviewed, which related to an outbreak of Escherichia coli O104 infection in May 2011 [22].Investigations of outbreaks in Germany were all led by the Robert Koch Institute.All investigations of outbreaks occurring in the Netherlands were supported or led by the Dutch National Institute for Public Health and Environment (RIVM).

Reporting of time intervals
Date of outbreak declaration was reported in 75 papers and at least one time interval was available from 55 of the 86 included papers.The hypothesis generation interval was more frequently available (50 papers) than the analytical study interval (28 papers).Of the 50 studies providing the hypothesis generation interval, 26 also included the analytical study period.Both intervals were available from nine of the 23 level 4 and 5 outbreaks and from 17 of the 63 level 3 outbreak investigations.A further two studies only reported the interval from hypothesis generation to availability of analytical results and three studies only reported the total time from outbreak declaration to availability of analytical results.

Time intervals by outbreak level
The median hypothesis generation and analytical study time intervals were shorter in level 4 and 5 outbreaks (median: 3 days; range:

Time intervals by study design
The most common study design was a case-control study (n=52 a Cross-regional or national impact with national coordination.
b National or international outbreak with potentially severe or catastrophic public health impact requiring national direction.
to availability of analytical results) was shorter in cohort studies than in case-control studies (Table 4).Unfortunately, however, the number of outbreaks reporting time intervals by study design was too small to make robust comparisons.
A small number of studies used other study designs, including case-case [44] and case series [16], or a combined approach of case-control and cohort [11,18,28,33,46,47].One level 4 outbreak used a mixed case-control and cohort design [47].This investigation was of a large-scale outbreak of E. coli infection with 135 linked cases and was completed around a week from outbreak declaration.The speed of this investigation is likely to have been aided by mandatory surveillance information gathered on cases in the two months before recognition of the scale of the outbreak, when a rise in the number of cases had been noted but no common source or E. coli subtype identified.Environmental samples were also pivotal in the testing of the hypothesis and prompt withdrawal of the implicated product from the market.

Time intervals by type of infection
The majority of papers (n=65) related to outbreaks of gastroenteritis.Where such outbreaks were suspected or known to be due to more severe infections (i.e.verotoxin-producing E. coli (VTEC), Shiga toxin-producing E. coli (STEC), Shigella sonnei), investigations appear to have been completed more rapidly than mild to moderate infections.The median investigation period (from outbreak declaration to availability of analytical results) for level 3 outbreaks of mild to moderate gastroenteritis was 31 days, compared with 10 days for level 4 and 5 outbreaks of severe gastroenteritis (Table 4).a Cross-regional or national public health impact with national coordination.b National or international outbreak with potentially severe or catastrophic public health impact requiring national direction.a Cross-regional or national public health impact with national coordination.b National or international outbreak with potentially severe or catastrophic public health impact requiring national direction.

Table 4
Time intervals of outbreak investigations by outbreak level in selected peer-reviewed publications containing epidemiological outbreak investigations (n=55)  b Cross-regional or national public health impact with national coordination.
c National or international outbreak with potentially severe or catastrophic public health impact requiring national direction.
The time it takes to complete an epidemiological investigation may in part be affected by the natural history of infectious diseases, namely the incubation period.
Where incubation periods are longer, it will take longer for cases to be detected, given the longer period until symptoms develop in exposed individuals, and there is therefore an increased risk of recall bias.This may lead to delays in the identification of the source of an outbreak and may present greater challenges in identifying additional linked cases, given the increased potential for the movement of cases.Unfortunately, the number of studies of non-gastroenteritis outbreaks in this review was relatively small (n=22) and with few of them reporting time intervals (n=9), it was difficult to identify patterns related to specific infections or incubation periods. In

Time intervals and number of cases
There was no correlation between the number of suspected and confirmed cases reported in an outbreak and the number of days from outbreak declaration to hypothesis generation (r 2 =0.0007).There was also no correlation between the number of cases and the number of days to completion an analytical report (r 2 =0.00001).

Discussion
This review found considerable variation in the speed of generating a hypothesis and obtaining analytical results following declaration of an outbreak with cross-regional, national or international public health impacts.The analytical study period following declaration of an outbreak tended to be shorter for outbreaks classified as level 4 or 5.This is likely to be due to a greater amount of resources being quickly mobilised following identification of an outbreak of this nature.By definition, such outbreaks are deemed to have the potential for severe or catastrophic public health impact and direction by national agencies will bring with it the ability to command greater resource deployment.
It should also be noted that the categorisation of outbreak level was based on definitions used by PHE and applied to outbreaks across Europe.Countries outside England may use different criteria to assess the potential impact of an outbreak, which may affect the level of response and timeliness of investigations.Therefore, conclusions regarding the timeliness of outbreak investigations by level should be drawn with caution.
The outbreak investigation period, which has a major influence on the timeliness of controlling an infectious disease outbreak, is just one part of a bigger picture.Whether an investigation period of 15 days for levels 4 and 5 and 31 days for level 3 from outbreak declaration to the availability of analytical results is acceptable needs to be considered alongside delays in outbreak recognition and notification, as well as how swiftly the required control measures are implemented following availability of the analytical results; all of which will affect the resulting population impact.
The study design appears to be one factor influencing the speed of completing analytical investigations; however, choice of study design is likely to be limited by the context of the outbreak.The longer time intervals in case-control studies may be due to challenges of identifying cases speedily and difficulties in selecting and recruiting appropriate controls.Hypothesis generation in cohort studies is likely to be swifter given the investigation starts with an identified population cohort.The factor that identifies the group as a cohort will itself provide clues to the source of the outbreak.In contrast, case-control studies often involve the identification of additional cases over time, which is likely to increase the time taken to define a hypothesis.While control selection methods for such studies was outside the focus of this review, some points are worth noting.Details of control selection methods were often lacking or sparse in the reviewed outbreak reports.Random digit dialling within specific postcode districts or other geographical areas was used in a number of investigations for which analytical results were available within three weeks of the outbreak being declared [6,20,25].In one investigation, which was completed in 22 days, cases were asked to nominate a number of controls (given relevant criteria) [9].A detailed description of control selection methods used by researchers at RIVM in the Netherlands was provided by Whelan et.al. [38] in a report of a level 3 case-control study, which had an analytical study period of 16 days (the median for this level in our analysis).RIVM receives each year a randomly selected list of 500 residents from each municipality (based on a unique reference number), totalling about 20,000 individuals per year.From this, a simple random sample of 300-500 individuals are invited to take part in an annual 36-question survey.The survey covers demographics, symptoms, travel history and risk factors in the previous 30 days.Completed surveys are used for enhanced surveillance of food-borne and respiratory infections and can be used in outbreak investigations, reducing the reliance on additional manpower for control selection and interviews [38].
The number of cases may be considered an important factor in the speed of completing epidemiological investigations, both in terms of the public health importance of a large outbreak necessitating a speedy response, and the amount of information available from trawling interviews on which to base hypotheses.However, no association was found between the number of cases detected and the speed of completing an epidemiological investigation among the studies included in this review.A longer delay in detecting an outbreak is likely to lead to a higher number of cases and a greater risk of an outbreak escalating to a higher level.Unfortunately, delayed recognition of outbreaks could not be analysed in this review as any difference between the date the number of cases in the population reached outbreak levels and the date an outbreak was recognised and declared by authorities was rarely reported.
There will be additional factors influencing the timeliness of analytical epidemiological investigations that were not available from the published reports in this review, such as the local public health systems, the availability of resources for investigations, the quality of surveillance and effective public communications.The severe acute respiratory syndrome (SARS) outbreak in 2003 heightened governmental awareness of the risks and impacts of international outbreaks in the context of increased global travel; the timeliness of investigations and public communications appears to have improved somewhat since [90].The use of electronic surveillance systems and algorithms to detect outbreaks has also improved the detection of outbreaks and the subsequent public health response in a number of countries [91][92][93].Such surveillance systems should be regularly evaluated in order to ensure their ongoing usefulness and contribution to timely outbreak investigations [93].In addition to effective, responsive surveillance systems, public communications regarding specific outbreaks can assist both in detecting outbreaks and in reducing the number of additional cases.
Publication bias is likely to have affected the findings of this study; not all languages were included in the review and there is likely to be a bias towards publishing reports where a source was identified.This may in part account for the large proportion of gastroenteritis outbreak investigations in this review; the short incubation times of these infections may reduce recall bias, leading to more reliable information on which to identify the source of the outbreak.There may also be some bias towards publication of swifter investigations that are considered unusual or highlight good practice.A previous review of published and unpublished food-borne gastroenteritis outbreaks found that few outbreaks reported to the Health Protection Agency (now Public Health England) led to peer-reviewed publication; those that were published had a bias towards more unusual outbreaks [94].Therefore, a review of time periods within unpublished outbreak investigations would complement this report.It would also be interesting for future studies to compare the timeliness of investigations with identification of a source and implementation of robust control measures.
Timeliness in outbreak investigations is important for minimising the number of people affected and protecting public health.Our findings provide a first overview of timeliness of analytical outbreak investigations.
Given the current lack of guidance, it will be useful to develop guidelines regarding what might be considered timely and how to improve the timeliness of outbreak investigations.
A key finding from this review is the need for more standardised reporting of time intervals in outbreak investigations so that the timeliness of investigations can be better understood.This will be required before firmer performance-monitoring standards can be developed.Our results suggest separate guidance and/or standards for the completion of analytical studies according to the severity of public health impact could be established.
While this review has provided useful material to inform discussions within PHE as to what might be considered as high quality in terms of timeliness, organisations within other countries may also find our results useful when considering factors influencing the speed of outbreak investigations and service improvements to ensure prompt completion of investigations.Separate recommendations for intervals from outbreak declaration to hypothesis generation and from hypothesis generation to completion of an epidemiological investigation could be considered; however, the number of studies reporting separate intervals is too small and the variation in timeliness too wide to draw firm conclusions from this review.The development of standards for performance monitoring requires further consideration.Such standards may assist investigation teams in getting organisational support for mobilisation of resources and lead to a more rapid public heath response.However, flexibility in applying standards to monitor the effectiveness of and improvements in outbreak investigations is advisable to allow for variation in the context and complexity of an outbreak.The introduction of such guidance or standards may usefully be accompanied by the development of tools to support prompt investigations.Robust routine surveillance systems and workforce capacity must also be maintained to enable prompt recognition of and response to outbreaks.
FigureTime intervals of outbreak investigations by outbreak level in selected peer-reviewed publications containing epidemiological outbreak investigations (n=55) Outbreak declaration to hypothesis generation Hypothesis generation to analytical resultsOutbreak declaration to analytical results

a
Median and range are not included where the number of papers reporting the time interval was less than three.

Table 1
Public Health England incident levels

Table 2 )
.A total of 19 outbreaks occurred across the UK (eight across two or more countries within the UK, four in Scotland, three in England, three in Wales and one in Northern Ireland).These were led by the Health Protection Agency (now Public Health England), Health Protection Scotland, the National Health Protection Service for Wales and the Communicable Disease Surveillance Centre for Northern Ireland, as appropriate.

Table 2
Number of selected papers, by country and outbreak level, in peer-reviewed publications containing epidemiological outbreak investigations (n=86) ).The proportion of case-control studies by outbreak level was similar in level 3 (38 of 63) and level 4 and 5 outbreaks (14 of 23).Approximately a third (n=18) of all case-control studies reported using matched controls.Cohort studies were carried out in 26 of the selected papers, with similar proportions across outbreak levels (19 of 63 level 3 and 7 of 23 levels 4 and 5).The median interval from outbreak declaration to hypothesis generation was shorter in level 4 and 5 outbreaks than in level 3 outbreaks, and the total investigation period (from outbreak investigation

Table 3
Time intervals of outbreak investigations by outbreak level in selected peer-reviewed publications containing epidemiological outbreak investigations (n=55) this review, level 4 and 5 outbreak investigations, with time intervals reported, included infections of STEC, VTEC, Shigella sonnei, Salmonella and hepatitis A. Level 3 reports, with time-intervals reported, [48]uded infections of E. coli, Cryptosporidium parvum, Campylobacter, Giardia lamblia, hepatitis A, Legionella pneumophila, Leptospira, norovirus, Pseudomonas aeruginosa, Salmonella, meningitis due to echovirus or coxsackievirus, Coxiella burnetii and Yersinia pseudotuberculosis.Only one report of a nosocomial outbreak was included in this review (due to P. aeruginosa infection)[48].This nosocomial outbreak took 68 days from outbreak declaration to form a hypothesis and no other time intervals were reported.