Clustering of diseases is known to occur since ages(1). Unexplained and unusual clustering of cases occasionally comes to notice of clinicians. Various theories have been proposed with regards to its etiology. Of many studies investigating close case aggregations of any type, a few have come with creditable explanation. The examples include the links found between angiosarcoma of liver and vinyl chloride monomer exposure(2) and mercury poisoning in Minamata, Japan(3).

Climatic or meteorological factors have been implicated in triggering the pathogenic mechanism. It has been suggested that a meticulous study of "combination of changing factors" may help find the underlying pathophysiology of the "puzzling clusters". In the present study we investigated some unexplained perceived clustering's, among hospital admissions.

The study was conducted among pediatric admissions in a tertiary care hospital of Central India, which has moderate climate. A one-year prospective study from 1st September 2002 to 31st August 2003 of all perceived clusters of disease or manifestations and 2 years retrospective analysis from September 2000 to August 2002 of significant clusters namely major nontraumatic bleeding manifestations and acute renal failure noted prospectively was done. Communicable disease clustering or clustering's where explanations were obvious or known has been excluded. Only admitted cases or patients in retention room for nonavailability of beds for more than one hour were studied. Patients who refused admission or shifted to other hospital were also excluded from study. Cases who were brought dead or in whom case definition criteria were not fulfilled were also excluded. Disease manifestations, day of onset, place of onset, detailed history, physical examination, investigations and treatment were recorded in all patients. The meteorological (climatic) database including daily (average of 4 values) and monthly average values of 4 parameters was obtained from the All India Institute of Meteorology, Pune (Maharashtra). The parameters recorded were maximum average temperature (T_max), minimum average temperature (T_min) in degree Celsius, average humidity (H), and average barometric pressure (BP).

The study was divided into 4 phases. In the first phase, all perceived unexplained clusters were recorded and studied. Because the total number of admissions will affect the number of cases, we worked with case rate, that is number of cases divided by total number of admissions. In second phase, falsely perceived clusters were excluded and true clusters were studied in detail. In the third phase, detailed retrospective analysis on some large significant prospective clusters was also performed. In the fourth phase, related literature was reviewed.

Relation of the clustering with factors like age, sex, climate, etiology, etc. was looked into using Chi-square test, Z test and Pearson’s correlation coefficient with SPSS software. Statistical analysis by centroid method was done using Cluster software.

The definitions, inclusion criteria for 8 identified clusters were as follows:

MNTBM and ARF case groups formed large clusters and 2 year retrospective study was carried out. BA, CM and AG case groups formed small cluster group of uncommon diseases, manifestations and studied separately. HBER, HCS and VUR case groups were falsely perceived clusters and excluded in the second phase of study.

The prospective study had 1744 total admissions with male:female (M:F) ratio 1:0.6. The cluster study group formed 11.98% of total admissions. The month wise distribution of cluster study cases in the prospective study is shown in Fig.1.It is seen thatMNTBMcases were clustered in October, November and December months, while ARF cases were clustered in March, April and May months.HBER, HS and VUR formed falsely per ceived clusters. BA, Ag and CM formed the rare clusters seen in October and November months. Relation of climatic factors with clustering of MNTBM and ARF over 3 years is shown in Fig.2. The case rate of MNTBM decreases with rising temperatures and vice versa whereas the case rate of ARF increases with rising temperature and vice versa which is seen consistently over three year period.

Fig. 1. Monthwise distribution of cluster study cases from September ’02 to August ’03

Fig. 2. Monthwise distribution of MNTBM and ARF case rates with temperature variation (September ’02 to August ’03).

In two-years retrospective and one-year prospective study of 5038 total admissions, MNTBM was noted in 288 cases (5.71% of total admissions) and ARF in 162 cases (3.22% of total admissions). Clustering of MNTBM was seen in months of October, November, December, and of ARF in months of March, April and May. These were statistically highly significant (X² = 15.52, P <0.005 and X² = 13.84, P <0.005 for MNTBM and ARF respectively) consistently over a 3-year period. Analysis by centroid cluster method revealed similar results.

Biliary atresia, acral gangrene and CM: Four cases of biliary atresia and 3 cases of acral gangrene were seen in month of November. 4 cases of CM were seen in months of October. All of them were male between 0-1 year age group. Being small clusters, no statistical tests could be applied.

In our study we had 8 perceived cluster study cases. Of these 5 were significant clusters namely major non-traumatic bleeding manifestations (MNTBM), acute renal failure (ARF), biliary atresia (BA), acral gangrene (AG) and cardiomyopathy of unknown origin (CM). MNTBM and ARF formed the major cluster cases, for which 2 years retrospective study was extended and BA, AG, CM formed minor cluster cases.

Three clusters namely vesicoureteric reflux (VUR), post-neonatal hypocalcemic seizures (HS) and hyperbilirubinemia in exchange range (HBER) were falsely perceived clusters as no statistical significance was noted on analysis and were, therefore, excluded from further investigations.

MNTBM, HBER, AG were manifestation clusters and rest disease clusters.

We had time clustering of MNTBM and ARF consistent over 3 years period during months of October, November and December (post-rainy season) and April, May, June (summer months) respectively. Small clusters of CM, BA and AG occurred in months of October and November.

The effect of climatic factors (meteorological factors) on MNTBM and ARF clusters has been investigated separately for prospective period and over entire 3 years period. Meteorological parameters vary each year; hence a prolonged study over several years is necessary.

Relationship between climatic (meteorological factors) and certain diseases like rheumatoid arthritis, acute gouty arthritis, SLE and Bechet’s disease is well documented. This is besides the infectious diseases which have a typical seasonal relationship.

Detailed search did not reveal any similar studies, but a few studies showed significant effect of climatic factors on incidence of intracranial bleeding during cold season.

Viral hemorrhagic fevers are known to form clusters; however, etiology analysis of our cases did not prove it as the reason for clustering. Only 4 cases of dengue hemorrhage fever were seen in cluster study group. Some studies mention unusual temporal clustering of bleeding disorders, viral hemorrhagic fever and sudden increase in non-specific syndrome as red flags of bioterrorism. Chyalte, et al. and Feiginul, et al. observed temporal clustering of bleeding in form of subarachnoid hemorrhage or intracranial bleeding during winter season(4,5). By contrast other studies showed clustering of subarachnoid hemorrhage in spring season(6-8), and autumn(8,9). Strangely females were more commonly affected by MNTBM (M: F ratio 1:2.1) during cluster months in our study despite the fact that males predominated amongst admission (M: F ratio 1: 0.67), which were consistent over 3 years. This higher susceptibility of females may have a genetic basis. Age group especially less than 1 year was significantly affected among cluster cases. ARF cases also clustered among females, children aged 0 - 1 years and during hot season which was consistent over 3 years. This suggests immature immune system and poor adaptive response to environmental changes or triggering cause clustering. Clustering of ARF during hot seasons suggest that dehydrating effect of the season especially on young has potentiated or enhanced expression of ARF.

Three rare clusters of BA, CM and AG occurred in months of Oct and Nov. The study group was not big enough to extract positive correlation. Such cases did not or only exceptionally occurred in rest of the months.

According to the latest literature BA can occur in some epidemics and clustering can be explained. Paula et al mentions it to be caused by environmental exposure(9,10). USA has also identified biliary atresia in seasonal clusters(11). Clustering of CM may be a result of cardiopathic viral infections during that season, however, we could not do viral studies to substantiate or mitigate the same. The clustering of AG remains unexplained.

We have taken manifestation clusters also and not disease clusters alone for investigating possible etiology or propounding factors. Manifestation analysis has been a primary tool of diagnosis and also of management for centuries in older medical systems like Ayurveda, Yunani, Chinese, Homeopathy, etc. The number of diseases defined only by manifestations are being identified more in modern medicine also e.g. fatigue syndrome, syndrome-X. Many patients with varied etiology presents with arthritis with seasonal variations(12-16). Manifestation analysis and management form a large part of clinicians practice and symptomatic and supportive therapy have great role to play. Under the circumstances, analysis of causative or potentiative factors of certain manifestations has its own importance.

The Center for Disease Control and prevention (CDC) USA has its own "Cluster Investigation Protocol". No such protocol is available in India and needs to be prepared. The cluster study also will increase the awareness amongst us in early identification of cases and treatment when clusters occur.

The study cases included disease and manifestation cases which were labeled as disease clusters and manifestation clusters respectively. Manifestation clusters mean aggregation of similar signs/symptoms i.e., common features whatever may be the etiology or disease. In other words, various disease or etiology can present with similar manifestations. The analysis of unexplained clusters of disease/manifestations will help health professionals in hospital and community surveillance. This method is a new tool, which can unearth epidemiological, ecological and other identified and unidentified causes of diseases or manifestations. The hospital staff will also be prepared to handle increased number of cases and will also be prepared for its treatment. Our aim shall be to make a protocol to investigate a cluster disease and also have criteria to stop investigating a cluster. Secondly, it will give a common theme of transparency of action and candid approach for investigation and hypothesis to test. A negative finding will be possible outcome for allaying concerns. The unusual temporal clustering of illness or a sudden increase in nonspecific illness or manifestations should be given the heightened state of alert and an appropriate time to review red flags for bioterrorism. Our knowledge of medicine has classified disease into category. The pathophysiological mechanism was well understood, when causative agent to most cases were found. It is a known fact that this arbitrary classification of pathology into disease or syndrome group in any field of medicine is not perfect. Hence, to discover new etiology or disease, we have to look into signs and symptoms and other associations.

This study has some limitations. Although case rate was used to eliminate the bias due to changes in total admissions, a change in admissions of any type of disease or condition can bring a change in case rate of the types investigated. Cluster significance was based only on one statistical analysis. Reconfirmation of cluster significance using another statistical method based on Euclidean distances to confirm a cluster would be more appropriate.

We conclude that the seasonal or temporal unexplained clustering of not only the diseases, but also the manifestations occurred and meteorological factors had a role in the possible causation. More factors may be operative and await study. Further prospective multicentric studies over long periods should be done on clusters for investigating influence of environmental and other factors on disease and manifestations.

Contributors: BS collected data, analyzed the data, performed literature review and drafted manuscript and will act as its guarantor for final approval. JS conceived and designed the study and critically reviewed the manuscript

Funding: None.

Competing Interests: None.

1. Smith D, Neutra R. Approaches to disease cluster investigations in a state health department. State Med 1993; 12: 1757- 1762.

2. Baxter PJ, Anthony PP. Angiosarcoma of the liver in Great Britain, 1963-73. BMJ 1977; 2: 919-921.

3. Tsuchiga K. The discovery of the causal agents of Minamata disease. Am J Indust Med 1992; 21: 275-280.

4. Chyalte D, Chen TL, Bronstein K, Brass LM. Seasonal fluctuation in the incidence of intracranial-aneurysm rupture and its relationship to changing climatic conditions. J Neurosurg 1994; 81: 525-530.

5. Paula WY, Joseph SB, Richard SO, Levy MJ, Main JK. Epidemiology of biliary atresia: A population-based study. Pediatrics 1997; 99: 376-382.

6. Feigin VL, Anderson CS, Anderson NE, Broad TB, Pledger MJ, Bonita R. Is there a temporal pattern in the occurrence of subarachnoid hemorrhage in the Southern Hemisphere? Pooled Data from 3 large population-based incidence studies in Australia, 1981 to 1997. Stroke 2001; 32: 613-619.

7. Rosenorn J, Ronde F, Eskesen V, Schmidt K: Seasonal variation of aneurysmal subarachnoid hemorrhage. Acta Neurochir (Wien), 1988; 93: 24-27.

8. Gallerani M, Portaluppi F, Miada G, Chieregato A, Calzolani F, Trapella G, et al. Circadian and circannual rhymicity in the occurrence of subarachnoid hemorrhage, Stroke 1996; 27: 1793-1797.

9. Ostbye T, LevyAR, Mayo NE. Hospitalization and case-fatality rates for subarachnoid hemorrhage in Canada from 1982 through 1991. The Canadian Collaboration Study Group of Stroke Hospitalizations. Stroke 1997; 28: 793- 798.

10. Davenport M, Dhawan A, Yoon P, Bresee JS, Oiney RS, Khoury MJ. Epidemiologic study of infants with biliary atresia. Pediatrics 1998; 101: 729-730.

11. Yoon PW, Bresee JS, Oiney RS, James LM, Khoury MJ. Epidemiology of biliary atresia: A population-based study. Pediatrics 1998; 101: 729-730.

12. Shutty MS, Cundiff J, DeGood DE. Pain complaint and the weather: weather sensitivity and symptom complaints in chronic pain patients. Pain 1992; 49: 199-204.

13. Redelmeier DA, Tuersky A. On the belief that arthritis pain is related to the weather. Proc Natl Acad Sci USA 1996; 93: 2895-2896.

14. Jamison RM, Anderson KO, Slater MA. Weather changes and pain: Perceived influence of local climate on pain complaint in chronic pain patients. Pain 1995; 61: 309-315.

15. Arber N, Vaturi M, Schapiro JM, Jelin N, Weinberger A. Effect of weather conditions on acute gouty arthritis. Scand J Rheumatol 1994; 23: 22-24.

16. Krause I, Shraga I, Molad Y, Guedj D, Weinberger A. Season of the year and activity of SLE and Behcet’s disease. Scand J Rheumatol 1997; 26: 435-439.