Risk of Community-Acquired Pneumococcal Bacteremia in Patients with Diabetes a Population-Based Case-Control Study

Pathophysiology/Complications ORIGINAL ARTICLE

Risk of Community-Acquired Pneumococcal Bacteremia in Patients With Diabetes A population-based case-control study

REIMAR WERNICH THOMSEN, MD1 SØREN PAASKE JOHNSEN, MD, PHD1 croangiopathy, and reduced lung func- HEIDI HOLMAGER HUNDBORG, MSC, PHD1 HENRIK CARL SCHØNHEYDER, MD, DMSC3 tion have been suggested to predispose HANS-HENRIK LERVANG, MD, PHD2 HENRIK TOFT SØRENSEN, MD, DMSC1 diabetic patients to lower respiratory tract infections (3). Recently, we showed that among patients with pneumococcal bacteremia, di- OBJECTIVE — We conducted this population-based case-control study to examine whether abetes was not associated with a higher diabetes is associated with an increased risk of community-acquired pneumococcal bacteremia. case-fatality (mortality rate ratio after 90 days ϭ 0.6 [95% CI 0.3–1.2]) (4). Data RESEARCH DESIGN AND METHODS — We included 598 cases in the North Jutland County Bacteremia Registry, Denmark, with residence in the county and a first hospitalization about diabetes and the risk of pneumo- for community-acquired pneumococcal bacteremia from 1992 through 2001. Ten sex- and coccal infection come primarily from case age-matched population control subjects per case were selected, using a unique personal iden- series. As reviewed by Smith and Poland tifier. Diabetes was determined by record linkage with the County Prescription Database (for (5), the reported prevalence of diabetes in prescriptions for antidiabetic drugs) and the Hospital Discharge Registry (for previous hospital- these series has varied from 1 to 20%, de- izations with diabetes or diabetic complications). We performed conditional logistic regression pending on the patients’ age, type of hos- to estimate odds ratios (ORs) for pneumococcal bacteremia among diabetic and nondiabetic pital, study period, and country, as well as persons, with adjustment for a range of comorbid diseases considered to be risk factors for methods for ascertainment of diabetes. It pneumococcal infection. is not clear whether any apparent associ- RESULTS — The crude OR for pneumococcal bacteremia in persons with diabetes was 1.9 ation between diabetes and pneumococ- (95% CI 1.4–2.6). After adjustment for comorbidity, the OR decreased to 1.5 (95% CI 1.1–2.0). cal bacteremia in the studies is causal or The impact of diabetes on the risk for pneumococcal bacteremia was most pronounced in adults related to the existence of confounding aged 40 years and younger (adjusted OR 4.2, 95% CI 1.1–16.7) and in persons without any other factors, since no control groups have been coexisting morbidity (adjusted OR 2.3, 95% CI 1.3–3.9). Under the assumptions that the used. Other diseases that have been sug- association was causal and that there is a 5% overall prevalence of diabetes in our study popu- gested as risk factors for pneumococcal lation, 24 of 1,000 admissions with incident pneumococcal bacteremia may be attributed to bacteremia are congestive heart failure, diabetes. chronic obstructive pulmonary disease, malignancies (pulmonary, hematological, CONCLUSIONS — Diabetes seems to be a risk factor for community-acquired pneumococcal bacteremia. and other), alcohol abuse, cerebrovascu- lar disease, liver cirrhosis, and HIV infec- Diabetes Care 27:1143–1147, 2004 tion (6–11). In a recent North American case- control study of 228 immunocompetent, t remains uncertain whether diabetes is pair a range of functions in neutrophils 18- to 64-year-old adults with invasive a risk factor for invasive pneumococcal and macrophages in vitro, including che- pneumococcal infection and 301 age- infection, as presumed in immuniza- motaxis, adherence, phagocytosis, and matched control subjects, the self- I reported occurrence of diabetes was 10% tion recommendations (1). Several bio- intracellular killing of microorganisms, logical mechanisms may contribute to an which may be important in limiting in- in cases and 4% in control subjects (odds increased risk of bacterial infection in di- vasion by bacteria in vivo (2). Further, ratio [OR] 2.5, 95% CI 1.2–5.1). How- abetic patients. Hyperglycemia can im- decreased immunity, pulmonary mi- ever, after adjusting diabetes for other ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● variables including race, sex, and comorbidity, the authors reported that the asso- 1 From the Department of Clinical Epidemiology, Aalborg Hospital and Aarhus University Hospital, Aalborg, ciation was no longer significant (risk Denmark; the 2Department of Endocrinology, Aalborg Hospital and Aarhus University Hospital, Aalborg, Denmark; and the 3Department of Clinical Microbiology, Aalborg Hospital and Aarhus University Hospital, estimates not given) (12). Thus, few if any Aalborg, Denmark. other population-based studies within a Address correspondence and reprint requests to Reimar Wernich Thomsen, MD, Department of Clinical proper epidemiological design exist about Epidemiology, Aalborg Hospital and Aarhus University Hospital, Stengade 10, 2nd floor, DK-9000 Aalborg, this issue. We therefore conducted a pop- Denmark. E-mail: [email protected]. Received for publication 28 November 2003 and accepted in revised form 6 February 2004. ulation-based case-control study in Den- Abbreviations: PAR, population-attributable risk. mark to examine whether patients with © 2004 by the American Diabetes Association. diabetes have an increased risk of com-

DIABETES CARE, VOLUME 27, NUMBER 5, MAY 2004 1143 Diabetes and risk of pneumococcal bacteremia munity-acquired pneumococcal bactere- Diabetes translated disease categories in the Charl- mia, as compared with persons without We identified diabetic patients among son index into corresponding ICD-8 and diabetes. case and control subjects by record- ICD-10 codes, similar to previous ap- linkage with two population-based regis- proaches, such as the translation to ICD-9 RESEARCH DESIGN AND tries as described previously (4): the codes by Deyo et al. (20). We identified METHODS — The study was con- North Jutland County Prescription Data- ICD codes for all previous hospitaliza- ducted during 1992–2001 in North Jut- base (17) and the County Hospital Dis- tions of case and control subjects in the land County, Denmark, within a charge Registry (18). We ascertained County Hospital Discharge Registry. To population of 496,000 inhabitants, ϳ9% presence of diabetes in the same way ensure an equal chance of being diag- of the total Danish population. The study among case and control subjects, namely nosed with comorbid diseases in bactere- population was homogeneous Caucasian by searching the databases for earlier mia case and control subjects, we only (96.7% inhabitants were of Danish origin hospitalizations with diabetes or earlier included diagnoses recorded before the by 1 January 1997) (13) and mixed rural prescriptions for insulin or an oral antidi- date of hospitalization of cases. The and urban. The entire population was abetic drug. We have recently estimated Charlson index was calculated, and three provided with tax-supported health care the predictive value of a diagnosis of dia- levels of comorbidity were defined: 0 by the National Health Service, allowing betes identified by this approach to be (“low”), corresponding to patients with free access to the county’s seven public 97% (95% CI 89–100%) (4). Information no recorded underlying diseases imple- on diabetes in the databases was recorded mented in the Charlson index; 1–2(“me- hospitals. Through the use of the 10-digit Ͼ civil registry number, which is unique to earlier than and independent of hospital- dium”); and 2(“high”). every Danish citizen and encodes sex and ization with bacteremia. To ensure that Since alcohol abuse is not included in date of birth, a complete hospitalization the classification of diabetes was indepen- the Charlson index, we collected data on and prescription history can be estab- dent from the case’s hospitalization, we alcohol-related disorders from the Dis- lished for each individual, and unambig- did not include 10 diabetic patients charge Registry (ICD-8 codes 291, 303, uous linkage between population-based among cases who in our prognostic co- 979, 980, and 577.10; ICD-10 codes F10, registries can be performed. hort study were diagnosed during the ad- K86.0, Z72.1, R78.0, and T51) in addi- mission with bacteremia. tion to the diagnoses included in the in- We classified diabetic patients as hav- dex. We further collected data from the Cases of pneumococcal bacteremia ing type 1 diabetes if they were aged up to Prescription Database on the use of anti- Patients older than 15 years with a first 40 years at diagnosis and were treated biotics and immunosuppressive therapy, hospitalization for community-acquired with insulin in monotherapy. Having type including corticosteroids before hospital- pneumococcal bacteremia were identified 2 diabetes was classified if they were ization, defined as redemption of at least in the population-based microbiological treated by diet alone or ever treated with one prescription for a systemic antibiotic County Bacteremia Registry, as described oral antidiabetics, or if they were older of any kind (Anatomical Therapeutical in detail earlier (4,14). We excluded 30 than 40 years at diagnosis, regardless of Chemical [ATC] classification system cases with bacteremia, compared with treatment. code J01) within half a year of admission, our previous cohort study, because they and redemption of at least one prescrip- were not residents in North Jutland Confounding factors tion for any immunosuppressive drug County; place of residence was a sampling To adjust for comorbid diseases that may (ATC codes L01, L04, and H02 AB) criterion in the present case-control both be risk factors for pneumococcal in- within 1 year of admission, respectively. study. fection and associated with presence of diabetes, we calculated a summary mea- Population control subjects sure of confounding due to comorbidities Statistical analysis Using the Central Population Registry, developed by Charlson et al. (19). The We used conditional logistic regression to which has electronic records on all Charlson index includes 19 major disease estimate ORs for community-acquired changes in vital status, including change categories, including the suggested risk pneumococcal bacteremia among dia- of address, date of emigration, and date of factors for pneumococcal bacteremia betic and nondiabetic patients, with asso- death for the entire Danish population mentioned, and has been adapted for ciated 95% CIs. Initially, we analyzed since 1968, we selected 10 control sub- use with hospital discharge registry data data by obtaining contingency tables for jects for each case individually matched in ICD databases (20, 21). We therefore the main study variables: pneumococcal by sex, age (same year of birth), and place considered it a suitable method to ad- bacteremia, diabetes, and the possible of residence (North Jutland County). The just for the patients’ overall level of confounding factors (i.e., all disease cate- control subjects were selected with the comorbidity. gories included in the Charlson index, the risk set sampling technique (15); that is, For calculating the Charlson index overall level of Charlson index, as well as the control subjects had to be alive and at score, a weight is assigned to each comor- alcohol-related disorders and the use of risk of first community-acquired pneu- bid disease category and the score is the antibiotics and immunosuppressive ther- mococcal bacteremia at the time the cor- sum of these weights. Diabetes was sepa- apy before hospitalization). We then esti- responding case was diagnosed. When rated from the Charlson index because it mated the OR adjusted for possible using risk set sampling, the estimated ex- was the exposure variable in this study. confounding factors. Stratified analyses posure OR in a case-control design is an All other diseases in the index were con- were performed according to sex, age- unbiased estimate of the relative risk (16). sidered as potential confounders. We first groups (Ͼ15–40 years, Ͼ40–65 years,

1144 DIABETES CARE, VOLUME 27, NUMBER 5, MAY 2004 Thomsen and Associates

Table 1—Characteristics of case subjects with community-acquired pneumococcal bacteremia for community-acquired pneumococcal and control subjects from North Jutland County, Denmark, 1992–2001 bacteremia according to presence or ab- sence of diabetes. The crude OR for pneu- Case subjects Control subjects mococcal bacteremia in people with diabetes was 1.9 (95% CI 1.4–2.6). After n 598 5,980 adjustment for comorbidity, the OR de- Diabetes creased to 1.5 (95% CI 1.1–2.0), indicat- Present 53 (8.9) 298 (5.0) ing that the association was confounded Not present 545 (91.1) 5,682 (95.0) by a higher level of comorbidity in the Age (years) 67 (18–94) 67 (17–94) diabetic group. When use of antibiotics Sex and immunosuppressive therapy before Male 283 (47.3) 2,830 (47.3) hospitalization was included in the anal- Female 315 (52.7) 3,150 (52.7) ysis, the adjusted OR remained un- Comorbidity* changed 1.5 (95% CI 1.1–2.0). Comorbidity index low (0) 312 (52.2) 4,386 (73.3) To evaluate the impact of diabetes on Comorbidity index medium (1–2) 219 (36.6) 1,316 (22.0) the risk of pneumococcal bacteremia in Comorbidity index high (Ͼ2) 67 (11.2) 278 (4.6) certain subgroups of persons, we strati- Alcohol-related disorders 26 (4.3) 72 (1.2) fied our analyses according to age-group, Data are n (%) or median (range). *Charlson index (see text). sex, and level of comorbidity (Table 3). Adults with diabetes Յ40 years of age were four times more likely to be hospi- Ͼ65–80 years, and Ͼ80 years), and level 18 to 94 years (median 67 years) (Table talized with pneumococcal bacteremia of comorbidity. 1). The most common focus of infection than persons of comparable age, sex, and To examine the impact of diabetes on was the respiratory tract (485 cases level of comorbidity without diabetes. At the overall risk of community-acquired [81%]), followed by the meninges (56 older ages, the adjusted ORs decreased pneumococcal bacteremia, we calculated cases [9%]). The focus was undetermined gradually. Compared with nondiabetic the population-attributable risk (PAR) for in 39 cases (7%). persons, diabetic individuals between 65 a diagnosis of diabetes (i.e., the propor- Table 1 shows further details about and 80 years (adjusted OR 1.3, 95% CI tion of all cases of community-acquired the 598 cases and 5,980 control subjects. 0.8–2.1) and more than 80 years (ad- pneumococcal bacteremia that are attrib- A total of 53 cases (8.9%) had either re- justed OR 1.2, 95% CI 0.6–2.2) had just utable to diabetes [16]). In case-control deemed a prescription for insulin or oral a slightly increased risk for pneumococcal studies, PAR may be calculated by the fol- antidiabetic drugs or had a discharge di- bacteremia. When we stratified according lowing equation: agnosis of diabetes recorded before the to the level of comorbidity, the associa- ϭ ϫ Ϫ date of hospitalization with bacteremia, as tion between diabetes and an increased PAR P (OR 1)/ compared with 298 control subjects {[p ϫ (OR Ϫ 1)] ϩ 1} risk of infection seemed to be largely re- (5.0%). According to our criteria, the vast stricted to persons without any comorbid The proportion of persons with dia- majority of diabetic subjects (340 of 351) diseases (adjusted OR 2.3, 95% CI 1.3– betes in our reference population, p, can had type 2 diabetes; therefore, we chose 3.9), whereas ORs in individuals with one be estimated by the diabetes prevalence in to consider diabetes as one entity in fur- or more other previously recorded diag- control subjects, since the population ther analyses. A considerably higher pro- noses were close to 1. Further, ORs ap- prevalence of diabetes is relatively low portion of case subjects than control peared to be higher in male than in female and our control subjects are representa- subjects (48 vs. 27%) had one or more diabetic individuals (Fig. 3). tive for all noncases in the population previously recorded discharge diagnoses Under the assumptions that the asso- (16). as evidenced in the Charlson index. ciation was causal and that there was an Statistical analyses were performed Table 2 gives crude and adjusted ORs overall prevalence of diabetes in the study with use of STATA software (version 8.0; population of 5.0%, the total PAR (etio- STATA, College Station, TX). The study logical fraction) was 2.4%. Thus, of 1,000 was conducted according to guidelines of Table 2—Crude and adjusted OR for commu- admissions with incident pneumococcal the regional scientific ethics committee nity-acquired pneumococcal bacteremia ac- bacteremia in our study population, 24 for use of clinical and laboratory data and cording to presence of diabetes may be attributed to diabetes. was approved by the Danish Registry Board (J.nr 2002-611-0060). Crude OR* Adjusted OR† CONCLUSIONS — We found a 1.5- Diabetes (95% CI) (95% CI) fold increased risk of community-acquired RESULTS — A total of 598 incident pneumococcal bacteremia in individuals cases with residence in North Jutland Not present 1.0 (ref.) 1.0 (ref.) with diabetes compared with individuals County and a first hospitalization for Present 1.9 (1.4–2.6) 1.5 (1.1–2.0) without diabetes. The impact of diabetes community-acquired pneumococcal bac- *Crude OR for presence of diabetes in cases with on the risk was most pronounced in pneumococcal bacteremia compared with sex- and teremia were identified during the study age-matched control subjects; †OR adjusted for younger adults, in persons without any period. Patients included 283 (47%) men level of comorbidity and alcohol-related disorders coexisting morbidity, and in males. and 315 (53%) women. Age ranged from (see text). The main strengths of our study are

DIABETES CARE, VOLUME 27, NUMBER 5, MAY 2004 1145 Diabetes and risk of pneumococcal bacteremia

Table 3—OR for community-acquired pneumococcal bacteremia according to presence of care (12). However, we were able to diabetes, stratified by age, sex, and level of comorbidity adjust for all major smoking-related diseases, which we consider proxy measures Crude OR* Adjusted OR† of smoking in our aged study population. (95% CI) (95% CI) Living with children attending daycare is probably an uncommon risk factor in per- Age (years) sons aged 50 years or more, who consti- Ͼ15–40 4.3 (1.1–16.6) 4.2 (1.1–16.7) tuted 73% of our study population. Ͼ40–65 3.2 (1.8–5.7) 2.1 (1.1–3.9) Observational studies suggest that Ͼ65–80 1.5 (0.9–2.5) 1.3 (0.8–2.1) pneumococcal vaccination reduces the Ͼ80 1.4 (0.8–2.6) 1.2 (0.6–2.2) incidence of invasive pneumococcal dis- Sex ease among adults and the immunocom- Male 2.2 (1.4–3.4) 1.8 (1.2–2.8) petent elderly (23). Therefore, we would Female 1.6 (1.0–2.5) 1.2 (0.8–2.0) like to emphasize that the general vaccine Comorbidity‡ coverage in proposed “at-risk” individuals Comorbidity index low (0) 2.3 (1.3–3.9) 2.3 (1.3–3.9) in our study population is much lower Comorbidity index medium (1–2) 0.8 (0.4–1.6) 0.8 (0.4–1.6) than that, for example, in many states of Comorbidity index high (Ͼ2) 1.1 (0.3–3.3) 1.1 (0.3–3.3) the U.S., where two-thirds of individuals *Crude OR for presence of diabetes in cases with pneumococcal bacteremia compared with sex- and over 65 years of age are vaccinated (24). age-matched control subjects; †OR adjusted for level of comorbidity (except when stratified by this variable) In our county, the uptake of pneumococ- and alcohol-related disorders (see text); ‡using the Charlson index (see text). cal vaccine unfortunately has been as low as 2 of 1,000 people per year since the uniformly organized medical health such as the study by Nuorti et al. (12). 1997 (25). Before 1997, when national care system, which allows for a truly pop- The positive predictive value of a diagno- vaccine recommendations were revised ulation-based design, and the ability to sis of diabetes established through the (Statens Serum Institut, Copenhagen, adjust for comorbid diseases. registries proved to be high. Concerning 1996), vaccine coverage rates were prob- Pneumococcal bacteremia is proba- completeness, Kristensen et al. (22) re- ably extremely low. Thus, we expect that bly a very common feature of pneumo- cently found that 76% of patients with pneumococcal vaccination did not have coccal pneumonia, but its detection is known diabetes could be identified by a a major impact on our estimates. Never- highly dependent on admission patterns county prescription registry similar to theless, diabetic individuals in our study and timing of blood sampling for culture. ours, as compared with an independent population may have been vaccinated at We cannot exclude the possibility that 34% by a regional hospital registry, by higher rates than nondiabetic persons. physicians caring for diabetic patients collecting data over a 1-year period only. The resulting bias, however, would lead may be more alert to possible infections. We thus find it likely that our combined to an underestimation of the true risk Thus, a higher proportion of cases of data sources are nearly complete re- for pneumococcal bacteremia in diabetic bacteremia may have been admitted to garding known diabetes. Both case and patients. hospital among patients with diabetes, control subjects may have included addi- To our knowledge, our study is the and blood cultures may have been or- tional patients with diabetes who have first population-based case-control study dered more frequently. Such surveillance never previously been hospitalized or to examine the risk of pneumococcal bac- bias would lead to an overestimation of drug treated, but we expect such misclas- teremia specifically in diabetic patients. the risk of pneumococcal bacteremia in sification to be nondifferential, leading to The North American case-control study diabetic patients. Nevertheless, we found a conservative risk estimate. (12) focused on cigarette smoking and in our previous prognosis study of the An increasing level of comorbidity did not have sufficient power to assess the cases included in this report that bactere- was associated with both presence of dia- relation between diabetes and invasive mia density on admission was similar in betes and the risk of pneumococcal bac- pneumococcal disease. Two earlier cross- diabetic and nondiabetic patients (44 vs. teremia and, therefore, confounded the sectional studies from institutions in 45% with low density) and that the me- association of diabetes per se with the risk North Carolina and Spain have compared dian level of C-reactive protein (277 vs. 204 in unadjusted analyses. Coding and diag- the prevalence of unspecified bacteremia mg/l) and the proportion with severe sep- nosis of comorbidity may have been more in hospitalized adults with and without sis (56 vs. 40%) was even higher among complete for diabetic patients due to diabetes (26,27). Both found a twofold diabetic patients, which argues against a more frequent hospitalizations. This increased prevalence in the diabetic more meticulous case-ascertainment would, however, lead to a conservative group. In the Spanish report, but not in among persons with diabetes. risk estimate. The same holds true for any the American study, the prevalence of Diabetes data in our study were col- nondifferential coding errors in the dis- bacteremia with S. pneumoniae was also lected prospectively and independently charge registry. A number of unmeasured increased in diabetic patients (13 of 5,667 of the patient’s hospitalization with pneu- factors may have had an impact on the hospital admissions compared with 97 of mococcal bacteremia. We thereby avoid- risk for pneumococcal bacteremia in this 95,725 hospital admissions of nondia- ed recall bias, which may hamper case- observational study, including tobacco betic patients); however, the figures were control studies where diabetes data are smoking, socioeconomic status, and liv- not adjusted for differences in age and based on interviews or questionnaires, ing with young children attending day- other risk factors.

1146 DIABETES CARE, VOLUME 27, NUMBER 5, MAY 2004 Thomsen and Associates

The specific biological mechanisms teremia: a 10-year population-based co- miologic Prescription Database of North linking diabetes with an increased risk of hort study. Diabetes Care 27:70–76, 2004 Jutland: a valid tool in pharmacoepide- pneumococcal bacteremia have not been 5. Smith SA, Poland GA: Use of influenza miological research. Int J Risk Safety Med established. Interestingly, in a recent and pneumococcal vaccines in people 10:203–205, 1997 case-control study from the U.K. that fo- with diabetes. Diabetes Care 23:95–108, 18. Andersen TF, Madsen M, Jørgensen J, 2000 Mellemkjær L, Olsen JH: The Danish Na- cused on genotypes, Roy et al. (28) re- 6. Afessa B, Greaves WL, Frederick WR: ported an increased risk of invasive tional Hospital Register: a valuable source Pneumococcal bacteremia in adults: a 14- of data for modern health sciences. Dan pneumococcal disease in patients with year experience in an inner-city university Med Bull 46:263–268, 1999 mannose-binding lectin deficiency. This hospital. Clin Infect Dis 21:345–351, 1995 19. Charlson ME, Pompei P, Ales KL, Mac- common immunodeficiency has been 7. Burman LA, Norrby R, Trollfors B: Inva- Kenzie CR: A new method of classifying associated with several autoimmune dis- sive pneumococcal infections: incidence, prognostic comorbidity in longitudinal eases, including systemic lupus erythem- predisposing factors, and prognosis. Rev studies: development and validation. atosus and rheumatoid arthritis, but to Infect Dis 7:133–142, 1985 J Chronic Dis 40:373–383, 1987 8. Gransden WR, Eykyn SJ, Phillips I: Pneu- our knowledge not with diabetes. 20. Deyo RA, Cherkin DC, Ciol MA: Adapt- mococcal bacteraemia: 325 episodes diag- In conclusion, we found a clearly in- ing a clinical comorbidity index for use nosed at St Thomas’s Hospital. Br Med J with ICD-9-CM administrative databases. creased risk of community-acquired (Clin Res Ed) 290:505–508, 1985 pneumococcal bacteremia in individuals 9. Kuikka A, Syrjanen J, Renkonen OV, J Clin Epidemiol 45:613–619, 1992 with diabetes, when compared with non- Valtonen VV: Pneumococcal bacteraemia 21. Halfon P, Eggli Y, van Melle G, Chevalier diabetic individuals. during a recent decade. J Infect 24:157– J, Wasserfallen JB, Burnand B: Measuring 168, 1992 potentially avoidable hospital readmis- 10. Watanakunakorn C, Greifenstein A, Stroh sions. J Clin Epidemiol 55:573–587, 2002 Acknowledgments— This study has re- K, Jarjoura DG, Blend D, Cugino A, Og- 22. Kristensen JK, Sandbæk A, Lassen JF, Bro ceived financial support from the Western nibene AJ: Pneumococcal bacteremia in F, Lauritzen T: Use and validation of pub- Danish Research Forum for Health Sciences three community teaching hospitals from lic data files for identification of the dia- (“Vestdansk Forskningsforum”) and by grants 1980 to 1989. Chest 103:1152–1156, 1993 betic population in a Danish county. Dan from the Medical Research Council of North 11. Yu VL, Chiou CC, Feldman C, Ortqvist A, Med Bull 48:33–37, 2001 Jutland, the North Jutland County Medical As- Rello J, Morris AJ, Baddour LM, Luna CM, 23. Dear K, Holden J, Andrews R, Tatham D: sociation, the “Heinrich Kopp’s Legat,” and the Snydman DR, Ip M, Ko WC, Chedid MB, Vaccines for preventing pneumococcal A.P. Møller Foundation for the Advancement Andremont A, Klugman KP: An interna- infection in adults. Cochrane Database Syst of Medical Science. Research at the Depart- tional prospective study of pneumococcal Rev CD000422, 2003 ment of Clinical Microbiology, Aalborg Hos- bacteremia: correlation with in vitro resis- 24. Public health and aging: influenza vacci- pital, is supported by a grant from “Det tance, antibiotics administered, and clin- nation coverage among adults aged Ն50 Obelske Familiefond.” ical outcome. Clin Infect Dis 37:230–237, years and pneumococcal vaccination cov- We thank the staff of the Hospital Discharge 2003 erage among adults aged Ն65 years: Registries in North Jutland County (Amtsgaar- 12. Nuorti JP, Butler JC, Farley MM, Harrison United States, 2002. MMWR Morb Mortal den) for preparing the data. LH, McGeer A, Kolczak MS, Breiman RF: Wkly Rep 52:987–992, 2003 Results of this study have been presented in Cigarette smoking and invasive pneumo- 25. Danish Medicines Agency: Lægemiddel- part at the 20th Annual Meeting of the Scan- coccal disease: Active Bacterial Core Sur- statistik, amtsfordelt: den primære sund- dinavian Society for Antimicrobial Chemo- veillance Team. N Engl J Med 342:681– hedssektor og sygehussektoren [article therapy, Odense, Denmark, October 24–26, 689, 2000 online]. Available from http://www. 2003. 13. Statistics Denmark: Population in Den- laegemiddelstyrelsen.dk/statistik/ mark: BEF1A: Population 1 January by aarsstatistik.asp. Accessed 11 August region, age, sex, marital status. Copen- 2003 References hagen, 2003 [article online]. Available from 26. Bryan CS, Reynolds KL, Metzger WT: 1. American Diabetes Association: Immuni- http://www.statistikbanken.dk/statbank5a/ Bacteremia in diabetic patients: compari- zation and the prevention of influenza default.asp?wϭ800. Accessed 27 October and pneumococcal disease in people with 2003 son of incidence and mortality with non- diabetes (Position Statement). Diabetes 14. Schønheyder HC, Højbjerg T: The impact diabetic patients. Diabetes Care 8:244– Care 23:109–111, 2000 of the first notification of positive blood 249, 1985 2. Johnston CLW: Infection and diabetes cultures on antibiotic therapy: a one-year 27. Carton JA, Maradona JA, Nuno FJ, mellitus. In Textbook of Diabetes. 2nd ed. survey. APMIS 103:37–44, 1995 Fernandez-Alvarez R, Perez-Gonzalez F, Pickup J, Williams G, Eds. Oxford, U.K. 15. Wacholder S, McLaughlin JK, Silverman Asensi V: Diabetes mellitus and bacterae- Blackwell Science, 1997, p. 1–14 DT, Mandel JS: Selection of controls in mia: a comparative study between dia- 3. Koziel H, Koziel MJ: Pulmonary compli- case-control studies. I. Principles. Am J betic and non-diabetic patients. Eur J Med cations of diabetes mellitus: pneumonia. Epidemiol 135:1019–1028, 1992 1:281–287, 1992 Infect Dis Clin North Am 9:65–96, 1995 16. Szklo M, Nieto FJ: Epidemiology: Beyond 28. Roy S, Hill AV, Knox K, Griffiths D, Crook 4. Thomsen RW, Hundborg HH, Lervang the Basics. Gaithersburg, MD, Aspen Pub- D: Research pointers: association of com- H-H, Johnsen SP, Sørensen HT, Schøn- lishers, 2000 mon genetic variant with susceptibility to heyder HC: Diabetes and outcome of 17. Nielsen GL, Sørensen HT, Zhou W, Stef- invasive pneumococcal disease. Br Med J community-acquired pneumococcal bac- fensen FH, Olsen J: The Pharmacoepide- 324:1369, 2002

DIABETES CARE, VOLUME 27, NUMBER 5, MAY 2004 1147