The frequency of hospitalization among the elderly in the United States caused by gastrointestinal diseases between 1991 and 2004 increased dramatically, especially hospitalization of elderly individuals with nonspecific diagnoses. We analyzed 6 640 304 gastrointestinal disease–associated hospitalization records in this 14-year period by comparing the peak times of nonspecific gastrointestinal diseases with those of specific diseases. We found that most nonspecific gastrointestinal diseases peak concurrently with viral enteritis, suggesting a lack of diagnostic testing for viruses, which may adversely affect the efficiency of prevention, surveillance, and treatment efforts.
The successful prevention of disease is best understood through the study of well-defined populations and outcomes. The aggregation of specific diseases, such as various gastrointestinal infections without diagnostic testing for specific causes, into nonspecific syndromic disease outcomes is common. This practice degrades the capacity to choose best preventive practices and eliminates the possibility of detecting newly emerging pathogens. The consequential public health implication can be more severe in vulnerable subpopulations such as the elderly, an immunologically weaker sector growing in both size and proportion in the United States.
Infectious diseases, including gastrointestinal infections, typically demonstrate seasonal patterns, suggesting similarities in etiological properties,1–3 dominant routes of transmission, and environmental determinants of these diseases.4,5 Comparing the seasonal patterns of nonspecific diseases with the patterns of known diseases may hint at the identity of nonspecific pathogens. We documented the seasonal patterns for hospitalizations that involved specific and nonspecific gastrointestinal conditions and compared the times at which their incidence peaked.
We abstracted hospitalization records from 1991 through 2004 from the Centers of Medicare and Medicaid Services data set, which contains records of all Medicare recipients and includes 93% to 96% of elderly individuals residing in the United States.6 Variables used included date of admission, age at admission, and up to 10 diagnoses based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).7 Cases were selected if any of the diagnoses fell under any of the 11 major categories and subcategories of gastrointestinal diseases: “intestinal infectious diseases” (ICD codes 001–009), “other and unspecified noninfectious gastroenteritis and colitis” (ICD 558.9), and “symptoms involving digestive system” (ICD 787).
We calculated category-specific hospitalization rates using annual population estimates obtained from the US Census Bureau as the denominator to adjust for changes in the elderly population. Hospitalization records for individuals older than 85 years were excluded from study because annual population estimates were not available for this age group.
We aggregated records into daily counts so that we could conduct seasonality analyses. The annual seasonal peak was determined with Poisson regression. Mathematical details are available elsewhere.8 Data analyses were performed with S-Plus 8.0 (TIBCO Software Inc., Somerville, MA). The statistical significance level was set at P < .05.
Patients aged 65 through 85 years accounted for 6 640 304 of the 8 640 522 records abstracted (76.8%). Hospitalizations associated with gastrointestinal illnesses increased 150.8% between 1991 and 2004 (from 260 565 to 653 653) whereas the US elderly population aged 65 to 85 years increased only 14.1% in the same period. Most of the records were related to symptoms involving the digestive system (ICD code 787; 61.2% of the selected records), followed by other and unspecified noninfectious gastroenteritis and colitis (ICD code 558.9; 21.7%) and infectious gastrointestinal disease and intestinal infections caused by other organisms (ICD code 008; 17.8%). Because the ICD 008 code includes pathologically diverse diseases, this category was divided into subcategories.
Figure 1 shows the counts, percentages, and seasonal patterns for the 11 outcome categories. Table 1 displays total counts, rates, estimated calendar year peak days, and the predicted peak period. All 3-digit major disease categories other than amebiasis (ICD code 006) and other protozoal intestinal diseases (ICD code 007) demonstrated significant seasonality (P < .05). The seasonal peaks of the 2 major nonspecific gastrointestinal conditions in the elderly (ICD codes 558.9 and 787) were temporally adjacent to that of viral enteritis (ICD code 008.6; 59th calendar day). The estimated peak times for ICD 558.9 and 787 were early March (58th and 61st calendar day, respectively; Table 1). The similar timing suggests that substantial numbers of hospitalizations coded as nonspecific, noninfectious enteritis or gastrointestinal symptoms are actually undiagnosed viral gastroenteritis.
Total Counts, Annual Hospitalization Rates, and Peak Times for Hospitalizations of the Elderly Associated With Gastrointestinal Infections and Symptoms: United States, 1991–2004
|Peak Timea (95% CI)||Predicted 2011 Peak Period||ICD Code||Condition||Countb||Annual Hospitalization Rate (per 1 million)||Seasonality Test P|
|20 (−43, 82)||…c||008.62||Adenovirus||31||0.06||NS|
|28 (20, 36)||1/20/11–2/5/11||008.63||Norwalk virus||204||0.41||<.001|
|31 (30, 32)||1/30/11–2/1/11||008.8||Other organism, not elsewhere classified||253 408||526.65||<.001|
|36 (27, 44)||1/27/11–2/13/11||008.69||Other viral enteritis||3560||7.34||<.001|
|55 (39, 70)||2/8/11–3/11/11||008.42||Pseudomonas||8761||18.21||<.001|
|55 (54, 56)||2/23/11–2/25/11||008||Intestinal infections caused by other organisms||1 183 148||2 439.88||<.001|
|57 (39, 76)||2/8/11–3/17/11||008.41||Staphylococcus||6248||12.91||<.001|
|58 (57, 59)||2/26/11–2/28/11||558.9||Other and unspecified noninfectious gastroenteritis and colitis||1 442 223||3 020.03||<.001|
|59 (55, 64)||2/24/11–3/5/11||008.6||Enteritis caused by specified virus||6412||13.27||<.001|
|61 (59, 64)||2/28/11–3/5/11||787||Symptoms involving digestive system||4 064 744||8 359.04||<.001|
|71 (70, 73)||3/11/11–3/14/11||008.45||Clostridium difficile||805 933||1 647.13||<.001|
|77 (76, 79)||3/17/11–3/20/11||008.4||Other specified bacteria||911 540||1 875.35||<.001|
|80 (75, 84)||3/16/11–3/25/11||008.61||Rotavirus||1828||3.75||<.001|
|139 (112, 166)||4/22/11–6/15/11||008.47||Other gram-negative bacteria||4294||9.05||<.001|
|146 (42, 249)||…c||006||Amebiasis||1304||2.72||NS|
|147 (134, 159)||5/14/11–6/8/11||005||Other food poisoning (bacterial)||17 816||36.86||<.001|
|157 (128, 186)||5/8/11–7/5/11||001||Cholera||738||1.53||<.001|
|167 (156, 178)||6/5/11–6/27/11||009||Ill-defined intestinal infections||63 007||130.40||<.001|
|168 (164, 172)||6/13/11–6/21/11||008.49||Other||67 275||148.10||<.001|
|169 (107, 232)||…c||007||Other protozoal intestinal diseases||5851||12.26||NS|
|173 (92, 253)||…c||008.67||Enterovirus not elsewhere classified||246||0.51||NS|
|181 (121, 241)||…c||008.3||Proteus (mirabilis) (morganii)||778||1.63||NS|
|185 (147, 222)||5/27/11–8/10/11||008.5||Bacterial enteritis, unspecified||7309||15.27||<.01|
|189 (185, 194)||7/4/11–7/13/11||008.43||Campylobacter||13 470||28.32||<.001|
|208 (139, 277)||…c||008.1||Arizona group of paracolon bacilli||23||0.05||NS|
|213 (190, 236)||7/9/11–8/24/11||002||Typhoid and paratyphoid fevers||490||1.03||<.001|
|225 (172, 278)||…c||008.2||Aerobacter aerogenes||385||0.80||NS|
|226 (218, 234)||8/6/11–8/22/11||008.0||Escherichia coli||5297||11.00||<.001|
|229 (−57, 515)||…c||008.44||Yersinia enterocolitica||311||0.65||NS|
|230 (219, 241)||8/7/11–8/29/11||004||Shigellosis||2820||5.95||<.001|
|233 (231, 235)||8/19/11–8/23/11||003||Other Salmonella infections||24 450||51.28||<.001|
|262 (135, 390)||…c||008.64||Other small round viruses||12||0.02||NS|
|296 (233, 359)||…c||008.65||Calcivirus||33||0.07||NS|
|318 (178, 459)||…c||008.66||Astrovirus||16||0.03||NS|
|364 (351, 376)||12/17/11–1/11/12||008.46||Other anaerobes||7989||16.60||<.001|
Note. CI = confidence interval; ICD = International Classification of Diseases; NS = nonsignificant. Values are arranged according to ascending peak day order. Details on derivation of peak times are available elsewhere.8
aIn Julian calendar days, where 1 represents January 1.
bTotal count from the 14-year study period.
cSeasonality not significant, and hence predicted peak not derived.
Our results show that the seasonal peaks for specific and nonspecific gastrointestinal infections occur at essentially the same time as enteritis caused by specified viruses. This finding suggests that some proportion of nonspecific infections could in fact have viral etiologies. Our temporal analysis, which involved high-resolution daily data spanning 14 years, independently supports the belief that there is a widespread lack of diagnostic testing and reporting.9,10 Elucidation of the true specific causes of these nonspecific conditions and incentives for diagnostic testing may improve the efficacy of prevention measures, population-wide surveillance, individual treatment, and detection of newly emerged pathogens. Although the estimated peaks for Staphylococcus and Pseudomonas infections also overlap the peak for diagnosed viral infections (as indicated by their overlapping confidence intervals), these organisms are rarely primary gastrointestinal pathogens and lack the biological plausibility of viral gastroenteritis.
The seasonality of infectious diseases is not yet fully understood. Two major driving forces, environmental and social, are most frequently suggested. Fluctuations in environmental parameters such as ambient and water temperatures,11–13 ultraviolet radiation,14–16 humidity or apparent temperature,17,18 and precipitation19 have been linked to infectious disease seasonality. With respect to social parameters, consumption of contaminated foods,20–23 exposure to contaminated environments,24,25 and institutionalization in a nursing home or hospital26,27 are common risk factors in the elderly.
Host biological factors, including declining immune function with age,28,29 disrupted or less than optimal intestinal microbiota,30,31 and previous exposure to antibiotics,32 are also known to modify susceptibility to and expression of gastrointestinal diseases. In addition, the manifestations of a gastrointestinal infection may involve other organ systems, often with a temporal delay, which may obscure the seasonal linkage. For instance, intestinal rotavirus infections appear to increase the risks of subsequent autoimmune diseases,33,34 and Campylobacter enteritis may cause 20% of all cases of Guillain-Barré syndrome, an acute flaccid paralysis.35
Moreover, there is increasing recognition that human and animal infections are frequently shared both geographically and temporally, suggesting that interdisciplinary “one health” approaches (holistic integration of human health, animal health, and ecosystem health) could further maximize disease prevention.36,37 Additional exploration of these factors, accompanied by etiological testing, may provide increased opportunities for the prevention and control of human disease.
This analysis involved several limitations. First, the observational nature of the analysis prevents causal inferences from being made. The increased frequency of nonspecific gastrointestinal illnesses could, in theory, indicate a new infection or infections, more known (but undiagnosed) infections, or a combination of both. Our reason for conducting this observational, exploratory study was that comprehensive clinical evidence on these observed synchronizations is lacking.
Second, we opted for analyzing elderly individuals aged 65 through 85 years instead of 65 years or older because of the lack of annual population estimates for individuals older than 85 years, which resulted in the exclusion of 23.2% of the total cases. The actual peak times could have been slightly biased had these individuals not been excluded from study. This older subgroup could have experienced the peak earlier (as a result of immunosenescence and a higher likelihood of being institutionalized) or later (as a result of decreased contacts with potentially infected people or other factors). Future analyses covering a more complete age range are recommended.
The efforts of public health professionals can be better optimized when the severity, spatial distribution, and temporal fluctuation of specific diseases are known.38,39 Results from time series analyses (e.g., predicted peak days) can be applied to optimize the timing of prevention measures such as vaccination. Recent successes with rotavirus vaccination in children illustrate the value of this approach.40 As we have illustrated, time series analyses may also have the potential to overcome some of the limitations imposed by nonspecific diagnostic coding and a lack of diagnostic testing by suggesting disease etiologies. However, we believe that a truly comprehensive understanding of the underlying seasonal associations between infectious disease exposures, outcomes, and modifiers will inevitably require more thorough diagnostic testing and research.
This work was supported by the National Institute of Environmental Health Sciences (grant R01 ES013171) and the National Institute of Allergy and Infectious Diseases (grants U19 AI062627 and NO1 A150032).
Human Participant Protection
The institutional review boards of the Tufts Medical Center and the Tufts University Health Sciences Campus approved the study protocol. No human participants were directly involved in the study.