Infectious-like spread of an agent leading to increased medical hospital admission in the North East Essex area of the East of England

This work documents the spread of a new type of infectious-like outbreak leading to a step-increase in emergency medical admissions. It employs a running 12 month sum of emergency medical admissions to detect step-like changes in admissions from small areas, namely Mid Super Output Area (MSOA) geographical areas and from GP practice enrolled patients. A general 25% step-like increase in emergency medical admissions occurred around November of 2008 in North East Essex due to gradual spatial spread of an infectious-like agent. Earliest incidence seems to be around August 2007 at a GP surgery in the Castle ward of Colchester leading to a 21% increase in medical admissions. The next major incidence appears to be around January of 2008 at a GP Surgery in the Shrub End ward of Colchester with a 32% increase in admissions. Sporadic occurrences can be seen in April and May of 2008 in four small areas with 21% to 65% increases in medical admissions. The major outbreak occurred in October, November and December of 2008, hence the apparent November 2008 date seen in the aggregated data for the entire area. Further sporadic spread continues through to October 2009. Several small areas appear to have escaped the outbreak, notably a rural GP practice in the village of Lawford and an area of Colchester dominated by student bed sits. More deprived elderly communities represent the social groups most affected. The outbreak shows strong age dependence which is reminiscent of what is termed ‘original antigenic sin’, i.e. age at first exposure to a strain of an infectious agent determines the quality of the immune response to later exposure to different strains of the same agent. There is evidence to suggest that residents of nursing homes are affected earlier and more strongly than others while those who die in hospital show only a modest increase, i.e. the infection generally leads to sustained poor health rather than death. Analysis of the primary diagnoses for those admitted to hospital strongly suggests that the agent may be the common herpes virus cytomegalovirus and this is consistent with the apparent time cascade in disease which emanates out of each outbreak. Other explanations may be possible. These findings have profound public health implications regarding the infectious origin of disease, to the funding formula used to distribute health care funds both in the UK and elsewhere and to the interpretation of age-standardized admission rates for medical admissions. Correspondence to: Rodney P Jones (PhD, ACMA, CGMA), Healthcare Analysis and Forecasting, Camberley, UK, Tel: +44(0)1276 21061; E-mail: hcaf_rod@yahoo.co.uk


Introduction
A series of long-term cycles appear to characterize the behavior of emergency medical admissions across the UK and elsewhere in the Western world [1][2][3]. Each cycle is composed of local infectious-like outbreaks which nest up over wider areas to give the appearance of a cycle at larger regional or national level. Each outbreak displays age and gender specific effects [4][5][6][7], and moves across the whole of the UK over a period of one to two years and are associated with sudden and unexpected increases in deaths, emergency department attendances and health care costs [3,4,[8][9][10] sufficient to create a long-term cycle in NHS surplus and deficit [11,12] and also appear to initiate a parallel cycle in the gender ratio at birth [13].
There are two suggested reasons for these infectious-like events. The first suggests that the events are merely the result of acute hospitals changing their threshold to admission [14], while the second suggests that this is a genuine infectious outbreak of a relatively slow to transmit persistent virus [1][2][3][4][5][6][7][8][9][10][11][12]. It does need to be pointed out that the first hypothesis cannot explain the increase in deaths or the cycle in the gender ratio at birth, however, it is possible to test these two hypotheses using the geographic area of a single hospital where the flow of patients is restricted to the hospital due to natural barriers to travel.
Such a situation exists in North East Essex, part of the larger East of England region. The population reside in three seaside townships of Harwich, Frinton-on-sea and Clacton-on-sea and in Colchester, a larger inland town which contains the main acute hospital. All roads converge on Colchester due natural barriers to travel provided by the estuary of the Clone to the south, estuary of the Stour to the north and the North Sea to the east. Surrounding countryside is largely dedicated to agriculture with a scattering of small villages and hamlets. Hence we have an ideal location in which to evaluate if hospital thresholds to admission or an infectious spread account for an unexpected increase in medical admissions which commenced in 2007 across the UK [3,[7][8][9].
In England all census data and resulting social classification of the population commences at the level of an output area (OA) containing roughly 300 head of population. Each OA is assigned to an output area classification (OAC) which consists of 7 supergroups, 21 groups, and 52 subgroups which area social classification constructed from 41 census variables. Each OA then nests up to a lower super output area (LSOA), to mid super output area (MSOA), to Parish, to Ward, to Local Authority and then to government or Health Service regions. Primary Care Organisations (PCOs) usually have a geography matching one or two Local Authority areas. Each Local Authority typically contains more than 100,000 head of population and in this respect NE Essex comprises two local authority areas of Colchester (population 170,000) and Tendring (150,000), serviced by 44 General Practitioner (GP) practices. Only 54% of the population lives in an urban area compared to 73% for England, and life expectancy at birth is only one year lower than the England average due to some pockets of high deprivation in some of the sea-side towns [15].
This study goes beyond the previous large area studies [1,[3][4][5][6][7]10], and moves the analysis of these events down to the small area of an English MSOA (roughly 5,000 head of population) to demonstrate small area spread consistent with an infectious event.

Methods
Data for emergency (non-elective) admissions was supplied by the former North East Essex Primary Care Trust (PCT) and covers monthly admissions for the residents of NE Essex over the period April 2005 to May 2010. Data did not contain patient identifiable features such as name, address, and postcode, date of birth or NHS number. Emergency (non-elective) admissions to the medical group of specialties were analyzed as a single cluster and include general medicine (42%), elderly medicine (33%), cardiology (9%), thoracic medicine (6%), gastroenterology (4%), oncology (3%), nephrology (1%) and other smaller medical specialties. Monthly trends for adult emergency medical admissions with a length of stay of one night or more were constructed and analyzed in a number of ways. In the first, all admissions at OA level were assigned into 17 concentric rings at 1 km intervals centered on the acute hospital in Colchester. The straight line distance of the population centroid of each OA to the hospital was calculated and this distance was used to allocate the admissions into the concentric rings. The aggregated data for each ring was available for analysis in this study. In the second method the admission trends in the largest MSOA were studied. In the third, all admissions were aggregated at the level of GP practice and trends for the largest practices were studied. Lastly data at OA level was aggregated to one of the 52 sub-groups in the Office of National Statistics (ONS) Output area classification (OAC) which are social groupings based on census data. The 2001 OA's and associated OAC's were obtained from the ONS. The OA code was then deleted and only the larger OAC aggregates were studied. None of the four methods involved patient identifiable aggregates, as discussed above, and the geographic units generally contained more than 100 admissions per annum to the wider medical group of specialties. Analysis of aggregates with less than 100 admissions is also not desirable from a statistical point of view since Poisson variation can act as a confounding factor in the analysis.
The potential contribution from Poisson variation to the value of any step-change was evaluated using Monte Carlo simulation of the ratio of two Poisson distributions. The 97.5% confidence Interval (CI) was calculated with 200,000 trials using Oracle Crystal Ball for an annual total (N) of between 100 and 700 in increments of 100. The resulting 7 values were plotted using Microsoft Excel and follow a power function where 97.5% CI = (1.965 × N -0.0891 ) -1 . This equation was then used to calculate the 97.5% CI associated with the step-increase observed in the various locations. Given the fact that a Poisson distribution becomes less skewed at higher numbers, when N> 1,000 then the 97.5% CI = 2.7 × n -0.5 .
One Figure in the results section presents actual monthly data which has been adjusted for the number of day per month and seasonal behavior. The seasonal (monthly) adjustment factors were calculated using the Solver function in Microsoft Excel to minimize the sum of the absolute difference between one month and the next in the adjusted monthly values.

Results
Trends were studied using running twelve month sums. Such a running twelve month sum avoids the need for complex analysis due to the seasonal nature of medical admissions. If demographic trends were the sole source of the admission behavior over time then such a running sum would generate a straight line whose slope would reflect approximately 1.0% to 1.5% per annum growth due to demographic shifts or what is known as the ageing population. In a running sum a step-change is transformed into a ramp. The base of the ramp marks the onset of the step-change and the ramp is only sustained if the stepchange endures for 12 or more months. Figure 1 presents the unusual situation confronted by the NE Essex PCT. Medical admissions had declined to a minimum by around March 2007, i.e. 12 months prior to March 2008 in the running sum trend, and continued at this low rate through to around November 2008 at which point there was a large and totally unanticipated 25% step increase resulting in serious financial pressure and a large ensuing deficit [16,17]. As can be seen there are no significant changes in any other specialty group other than general and elderly medicine and this argues against a hospital-wide reduction in the admission threshold or even a reduction in the threshold for admission through the emergency department. While the boundary between general and elderly medicine is somewhat fluid the average age is higher in elderly medicine. Note the slightly different time profiles between the two, especially the earlier larger step-down in elderly medicine. To give a statistical context to each of the trend lines the maximum possible Poisson-based variation is as follows (99.9% CI as a number and percentage): Trauma (±116, ±8%), All Others (±132, ±7%), Pediatrics (±141, ±7%), Surgery (±172, ±6%), General Medicine (±342, ±3%), Geriatric Medicine (±269, ±4%). While these statistical fluctuations can explain some of the small undulations seen in the non-medical specialty trends they are totally unable to explain the behavior in General and Elderly Medicine. To gain further evidence for an infectious-like spread Table 2 summarizes the details for the largest MSOA aggregates. Note that in these smaller geographical units the initiation dates show a wider spread from April 2008 to October 2009 while the increase ranges from +16% to +70%. In this table, the 97.5% confidence interval (CI) has been calculated from Poisson variation and represents the maximum possible contribution from Poisson variation to the step-change. Hence for the first MSOA (E02004517) in Table 2 the measured 65% step-increase has a maximum possible 15% increase which could be due to chance, i.e. something higher than a 50% step-increase can be guaranteed to be due to the infectious agent, on the other hand the real step change could be as high as 80%. Even the smallest MSOA (Colchester 010A) with only 131 admissions in the twelve months prior to the step-increase has a change well beyond the 97.5% CI for Poisson variation.
Since GP practices service their patients from the close geographic vicinity Table 3 investigates the impact at GP practice level. An even wider spread in the initiation date is detected using this method from August 2007 through to June 2009 with the usual wide range in magnitude of the step change. The point of great interest is a single practice in the Lawford area which experiences no apparent increase in medical admissions over the period studied. Lawford is a rural village in the Tendring LA some 15 km to the northeast of Colchester, and is situated in an otherwise sparsely populated area. Figure 3 shows the actual monthly medical group admissions (rather than a running sum) in the concentric rings with onset clustered around October, November and December 2008. This data has been  divided by the total days per month to give admissions per day and seasonal behavior has been adjusted for by applying monthly adjusting factors as described in the methods. The large step-change can be clearly seen, i.e. whatever has happened has reached near maximum impact within the space of a month.
The same set of rings were then used to see if age had any effect on the magnitude of the increase and this is presented in Figure 4 where there appears to be a general declining impact between ages 16 to 30 and a general very high impact above age 96. More curiously the effect of age appears to follow a series of peaks and troughs with some ages, most notably, 75, 79, 92 and 69 showing very small increases. Table 4 then investigates the possibility that particular social groups may be more affected. As can be seen the social group least affected live in an area dominated by student bed sits which experiences no increase in medical admissions, followed by areas of rural light industry and younger affluent families. The areas' most greatly affected contain higher proportions of elderly especially blue collar workers and more deprived residents living in public housing. Figure 5 investigates if the destination at discharge has any impact on timing or extent. As can be seen those who are characterized by the need for a care/nursing home show initiation as early as June 2008 and have a very high increase in admissions which does however tail off,  Table 3. Characteristics of the step-like increases observed for the largest GP practices.
presumably due to higher numbers of deaths. Those who die in hospital seem to initiate around November but show a lesser increase than those who are discharged back to their home or usual place of residence.
Admissions from (and returning to) prisons, psychiatric institutions and facilities for those with learning disabilities all show no change (data not shown), i.e. are not affected by the outbreak. These sources of admission represent a segment of the population having least contact with the rest of the general population.
Finally Table 5 investigates if the step-increase is specific to particular body systems. If this is a true infectious outbreak then pockets of diagnoses could be hidden across all specialties. In this table emergency admissions to any specialty and with any length of stay have been analyzed by looking at the ICD-10 chapter of the primary diagnosis for the admission. As can be seen the effects of the outbreak appear to stretch across multiple body systems. Further analysis at ICD 3 digit level of the 100 highest volume diagnoses shows that some 53

OAC Description
Step  diagnoses with 6,661 admissions per annum can be positively excluded compared to 1,602 admissions in 44 diagnoses which are strongly affected (data not shown), i.e. there are condition specific effects as well as conditions which remain unaffected. Why this could be possible will be explored in the discussion.

Discussion
In England, Primary Care Organizations (PCO) pay acute hospitals on a cost-per-case basis using the NHS tariff for services called the Health Resource Group (HRG) Tariff. Medical group emergency admissions account for around 18% of the entire inpatient volume (including day case, mental health and maternity) and hence a 24% increase in medical admissions will lead to an approximate 4.4% increase in the total inpatient costs. Such a large and unexpected increase will be financially crippling and this was indeed the case in NE Essex [16,17]. These outbreaks occur across the whole of the UK and it has been estimated that total costs rose by a minimum of £600 million following the 2002 and 2007 outbreaks [4,8,12,18] Figure 5. Effect of discharge destination on the time of initiation and extent of the step-like increase as revealed using running 12 month totals.

Chapter ICD Chapter for primary diagnosis
Step Change (%) Initiate Admissions in 209/10 (n) Step-change (n) *Emergency admissions to any specialty, and including zero day stay admissions. The bulk of the analysis presented in this study is for overnight stay emergency admissions, i.e. same day or zero day stay admissions have been excluded. There are several reasons for this, namely, zero day stay admissions fall into a grey area between genuine inpatient and emergency department care. This grey area is further complicated by the arbitrary four hour target for treatment in the emergency department in England which, since introduction in 2004, led to a large increase in zero day stay admissions as hospitals opened medical assessment units (MAU) [20,21]. While MAUs are accepted good practice they are also outside of the scope of the four hour target, and hence the validity of such 'admissions' are sometimes questionable [20,21].While this should not be a major issue for admissions to a single hospital the analysis has, never the less, been restricted to those patients with a length of stay encompassing one or more nights and represent those with a genuine need for hospital admission.
While the first method of analysis using concentric rings may be considered arbitrary, the geography of NE Essex is suited to this approach in that the first five rings dissect the town of Colchester, the next rings dissect various farming villages and hamlets and the final rings dissect the three coastal towns. Indeed whatever the method when dealing with a change in admission threshold for a single hospital admissions from any location should show a simultaneous jump which as Figure 1 demonstrates is obviously not the case. That the dates for initiation within the concentric rings are so widely disparate strongly suggests that the hospital itself is not a common focal point for spread of the infection. Indeed the initiation surrounding GP practices in Table 3 is suggestive that these are the most likely focus for infectious spread.
In a running sum, a permanent or semi-permanent step change in admissions due to a reduction in admission thresholds leading to more admissions or due to the introduction of a persistent infectious agent will create a ramp increase which lasts for 12 months. The foot of the ramp marks the point at which the permanent change has been made and the point twelve months on provides the magnitude of the initial step-change. Note that the step-change has to endure for 12 or more months in order to generate the ramp. If the step-change endures for longer than 12 months the ramp is then followed by a plateau as seen in the 3 km and 5 km concentric rings in Figure 2.
When dealing with such a curious phenomenon it is always useful to look back to see if it has occurred before. Indeed such step-like increases which are specific to medical admissions, i.e. such trends are not observed in trauma or in surgical specialties, have occurred before in the UK and have been documented around 1993,1996,2002,2007 and again in 2012 [1][2][3][4][5][6][7][8][9][10]22]. Even more curiously initiation always appears to commence in Scotland earlier than the rest of the UK [3], and each event not only results in an increase in medical admissions but an increase in deaths, a somewhat non-specific and more variable increase in GP referral and a general increase in health care costs [1][2][3][4][5][6][7][8][9][10][11]22].
The curious step-down prior to the step-increase has also been documented for emergency medical admissions in Northern Ireland [7], for an associated long-term cycle in deaths [3,22], and has been commonly observed by the author in other locations (unpublished). If this is not an acute hospital admission threshold phenomena then what are the alternatives? Firstly, how do we explain the step down which appears to occur elsewhere and also with deaths (all-cause mortality). While 50% of deaths may take place in hospital it is hard to conclude that a tightening of the admission threshold would lead to a reduction in all-cause mortality since the two are largely unrelated. Indeed most hospital deaths are to do with end of life care or, more correctly, the absence of alternative options for end of life care, i.e. the person will die with or without hospital care [10,23].
If the cluster of locations with high admissions around November and December of 2008 explain the overall trend observed in Figure  1, i.e. the high level trend is driven by a mass effect from the largest locations, then Figure 2 gives a clear indication of how an apparent step-down can likewise occur across a wider geography in that the 22 km ring has commenced a decline some 12 months after the onset of the step-increase. This is also seen in Figure 5 for discharge to a nursing home where there will be high turn-over of residents due to death. Hence the initial step down seen in Figure 1 is a reflection of the tailend of events following the 2002 outbreak, which similar to the 2007 outbreak may also have occurred later in NE Essex than across other parts of England.
If the step-down arises from the eventual loss of infectious potency, i.e. it is eventually bought under immune control, then the step-up must also be explained as part of an infectious spread moving across the whole of the UK along with a step increase in other factors such as GP referrals [24] which have nothing to do with acute admission thresholds. Indeed the highly variable initiation dates ranging from June 2008 through to June 2009 within the concentric rings and other modes of analysis surrounding the acute site cannot be explained by a common acute threshold which would equally apply across all locations at the same time. This is further reinforced by the variable magnitude of the step-change from no change in a minority of small area locations (Table 2) through to very large increases in a minority of other locations.
All analysis presented in the Tables and Figures contradicts the hospital admission threshold hypothesis. My own discussion with hospital Chief Executives and Managers leads to the clear conclusion that the threshold to admission is a clinical decision into which managers do not seek to intrude, other than by the provision of medical assessment units, etc. Research conducted in the USA demonstrates that this clinical threshold is maintained despite considerable fluctuation in the number of arriving admissions and that clinicians manage the resulting bed pressures by flexing the threshold to discharge rather than admission [25].
It is interesting to note that the NE Essex PCT had assumed that the step-down was due to their own efforts and were very proud of the fact that they had the lowest level of age-standardized medical admissions in the East of England. Subsequent analysis revealed that the step-up had merely occurred earlier in other locations in the East of England [16,17], leaving the PCO with the incorrect perception that they had done something special.
In the UK, GP practices enroll patients from a defined area and the only exception is for patients who subsequently move and wish to stay with the practice. In NE Essex GP practices therefore have patients from 9 to 37 MSOA (23 median) and one MSOA (E02004581) has admissions from 32 practices, however, the largest practices have 20% to 50% of admissions from just one MSOA rising to 90% in some of the smaller practices. Hence interpreting data at practice level reflects the combined effect of varying degrees of distributed populations, and once again the overall trend will be largely driven by the one or two MSOA with the largest number of admissions.
It has been suggested that the infectious agent is most probably cytomegalovirus (CMV). CMV is a ubiquitous herpes virus with a multitude of genes dedicated to immune evasion and subversion [1][2]26] CMV has been implicated in autoimmune disease [26][27][28] as an oncomodulatory and oncogenic agent [29][30][31][32] and is associated with very high mortality rates especially in those with high levels of anti-CMV antibodies and/or with an associated inflammatory response and in the elderly [33][34][35][36][37][38].
Do the broad effects against almost every body system shown in Table 5 support the CMV hypothesis? There are several key observations which suggest that this could be true.
1. CMV is known to increase all-cause mortality, i.e. the cause of death covers all ICD chapters. One study among elderly Latinos demonstrated 43% higher all-cause mortality in those with CMV IgG antibody levels in the upper quartile range while this group also had 35% higher mortality from cardiovascular disease [34]. Another study in Norfolk, England demonstrated 23% higher all-cause mortality in the high IgG antibody group with a 24% increase in death due to cardiovascular disease, a 13% increase in death due to cancer, and a 35% increase for the remaining causes of death, of which 12% had respiratory diseases, 16% gastrointestinal and 21% nervous system disorders [35]. Another study on the elderly in Nottingham, England calculated a sub-hazard ratio 94% higher for cardiovascular death for those with CMV infection and 21% higher for respiratory death [37]. A study comprising a representative sample of US nationalities demonstrated a 30% increase in both all-cause and cardiovascular mortality in CMV seropositive individuals with high C-reactive protein (CRP) levels [33]. Based on the simple fact that nearly 50% of people die in hospital [23], and that the bulk of a persons' lifetime hospital admissions occur in the last six months of life [23], the fact that CMV is so widely implicated in all-cause mortality would suggest that a CMV outbreak would therefore affect admissions for a wide range of diagnoses.
2. CMV is known to have potent effects on immune function affecting both autoimmune functions [26][27][28], and inflammatory pathways [28,[32][33][34][35][36][37]. There are over 100 known auto-immune diseases and over 100 inflammatory diseases and so the potential to predispose individuals to a wide range of such diseases leading to increased hospital admission in the face of a CMV outbreak is likewise feasible.

Systematic reviews of hospital admissions and death where
CMV was the confirmed causative agent in non-immunosuppressed patients demonstrated that the gastrointestinal tract and central nervous system were the most frequent sites for severe infection, while other common organ-systems included hematological disorders, thrombosis, ocular and lung disease [28,38]. Once again this list is consistent with Table 5 and the additional analysis of individual diagnoses.

A large increase in admissions for Signs and Symptoms, i.e.
where the diagnosis is somewhat vague, has been noted to commonly accompany these outbreaks [1,[5][6][7], and given the general lack of awareness for the wider role of CMV in hospitalization and death [1,2] many of these admissions are likely to be CMV-related but will lack appropriate diagnostic tests to confirm the real source. 5. CMV is known to infect the thymus with resulting tissue damage and thymopoiesis [1,2]. Recent research has demonstrated that thymopoiesis is associated with a systemic inflammatory state in the elderly [39], and that both thymic output and level of CRP are independent predictors of the time to end of life [40].
In this respect the list of diagnoses identified as possible candidates for a step increase appear to have a CMV link. Links with type I and II diabetes have been implicated in a number of studies [32,[41][42][43][44], and the increase in admissions for both type of diabetes suggests an agent capable of exacerbating such conditions in a clinical context. Evidence for active CMV infection is common in patients with liver cirrhosis 45 and once again we have an agent capable of at least exacerbating the condition in patients. Epilepsy and Intestinal disorders are characteristic diagnoses identified for these outbreaks [1][2]28], and in the individual diagnoses noted for NE Essex there were additional clusters of cardiovascular (angina, myocardial infarction, ischemic disease, atrial fibrillation and intercerebral hemorrhage), intestinal (diverticular disease, intestinal disorders, alcoholic liver disease, pancreatitis and other digestive disorders), lung (pneumonia unspecified, asthma, pleural effusion), and a nest of signs and symptoms (pain in throat and chest, abdominal pain, nausea, syncope, convulsions) all of which are highly indicative of CMV [28].
Also implied in the results of Table 5 is the concept of a disease progression as a time cascade. That such a time-dependent cascade could exist has been inferred from a study of GP referral to different specialties across the UK following the 2007 outbreak [12], from the peak in outpatient attendance for various dermatological conditions [46], and from the time trends for first diagnosis of particular cancers in the USA [47]. Such a time cascade would arise out of the timedependent effects upon the immune system mediated inflammatory and auto-immune functions. These preliminary results will hopefully act as the catalyst to further research.
A further interesting issue relates to the apparent saw tooth behavior seen in Figure 4 for the relative increase in medical admissions and age. Such a saw tooth pattern is typical of that seen in what has been called 'original antigenic sin' [48,49]. This arises when the body is exposed to the first of a number of different strains of the same agent. The immune response is 'primed' to the first strain and this priming may help or hinder in subsequent exposure to different strains, hence the series of peaks and troughs which depend on the timing of different outbreaks and the number of outbreaks to which individuals have been exposed.
While the public health implications of this study should be widely apparent there are equally profound implications to funding and the use of age-standardization. These will now be discussed.
Almost all countries use some form of capitation formula to distribute health care funds to various regions and local areas. In the USA such a formula is used to distribute Medicare and Medicaid funds to the various states. All of these formula are based on the assumption that costs are fundamentally driven by population characteristics such as age, gender and deprivation [50]. This work elegantly demonstrates that the medical-related cost of healthcare is modified by the pattern of infectious spread associated with these outbreaks and imply that genuine fair funding requires some form of retrospective adjustment to account for the timing and extent of the outbreak [50,51].
Age-standardization of admission rates relies upon two assumptions, namely, that the most commonly used five year age bands are widely relevant, and the rates within the bands are fairly stable over time. This is called the constant rate fallacy [52]. The very high degree of age-specific behavior which presumably arises from the outcome of 'original antigenic sin' elegantly demonstrates that such age-standardization is highly questionable when applied to the medical admissions so dramatically affected by these outbreaks. This probably explains why age standardization performs so poorly when attempting to forecast future numbers of medical admissions [53], simply because it is the step-like increases which drive the long-term trend, and not the underlying population demography.
The final issue to be addressed is the unique kinetic characteristics of these outbreaks. Fairly slow spread across the entire UK over an 18 month to two year period has been noted for both the 2012, 2007 and previous outbreaks and this corresponds with the relatively difficult to transmit nature of CMV [54] which would imply the introduction of a new strain of CMV [1,2] as per the age-specific behavior suggestive of antigenic original sin. CMV super infection with multiple strains is widely recognized as leading to more averse clinical outcomes [1,2] as implied by this study. How does this reconcile with the apparently rapid spread at local level such that within the space of around one month admissions jump from a lower to higher state? Like all persistent infections CMV is in a state of continuous outbreak and in many instances expresses itself in an acute influenza-like illness [1,2] and associated respiratory illness [55], which in children is statistically more frequent for those infected with CMV [56]. Shedding of the virus is common in both children and the elderly [57][58][59][60][61]. CMV has been detected in the air surrounding patients with CMV pneumonia and to a lesser extent for patients with a 'latent' infection [62], and CMV is known to survive for up to six hours on rubber, cloth and food and has been found on 83/90 wet and 5/40 dry surfaces in day care centers for children [63]. Studies have shown that on average an infected person transmits CMV to two susceptible people [61], and a combination of the above is likely to explain the mini-epidemics of CMV illness described in a neighborhood, renal transplant unit, a laboratory, in burns units and an elderly care ward [64][65][66][67][68]. The particular nature of the infectious spread for nursing homes and GP practices identified in this research strongly suggest that both of these may be acting as loci for the bursts of rapid epidemic-like local spread.

Conclusions
A phenomenon with infectious-like spread has been shown to dominate the time-dependent admission trends for specific medical conditions subsequent to a UK wide infectious outbreak commencing in 2007. This spread has all the characteristics of a genuine infection with a high degree of age specificity, spatial granularity, higher impact on locations with higher density of the elderly, especially the more socially deprived or to nursing home residents. The pattern of conditions most affected points to the involvement of CMV although this requires confirmation. Given the huge financial impact of these outbreaks [4,18,69] research is urgently required to establish whether CMV is the ultimate cause and hence what public health measures can be implemented. Indeed these results add weight to the infectious basis for many diseases [70][71][72], although genetic and lifestyle factors are also important. There are profound effects upon health care finances, the equitable allocation of funds and the calculation of health insurance premiums. There are additional wider implications to age standardization of admission rates and to standardized hospital mortality rates.

Postscript
This paper was initially peer reviewed and accepted for publication in the 2014 edition of 'Biomedicine International'. This journal ceased publishing in early 2014, and the original publication has now been kindly published by FGNAMB. Since 2014 additional small areas studies have been conducted confirming the results in this study. One of these studies has been published in FGNAMB 1(2): 42-54. These and other papers on these infectious-like outbreaks can be most easily located using a search engine such as Google Scholar.