Improving Patient Care Annals of Internal Medicine Risk for Hospital Contact with Infection in Patients with Splenectomy a Population-Based Cohort Study Reimar W

Improving Patient Care Annals of Internal Medicine Risk for Hospital Contact With Infection in Patients With Splenectomy A Population-Based Cohort Study Reimar W. Thomsen, MD, PhD; W. Marieke Schoonen, MSc, PhD; Do´ ra K. Farkas, MSc; Anders Riis, MSc; Jacob Jacobsen, MSc; Jon P. Fryzek, PhD; and Henrik Toft Sørensen, MD, DMSc

Background: Splenectomy has been associated with increased risk hazard of infection was 4.6-fold (CI, 3.8 to 5.5) higher in splenec- for infection. tomized patients than in general population comparisons from 91 to 365 days after splenectomy and 2.5 times (CI, 2.2 to 2.8) higher Objective: To assess the magnitude and duration of risk for hos- more than 365 days after splenectomy. The risks were similar for pital contact with infection associated with splenectomy. pneumonia and were higher for bacteremia. Markedly increased Design: Population-based cohort study. risks were also found when compared with those of appendectomized patients. Modest increases in infection risk were seen with Setting: Denmark. splenectomy matched-indication comparisons (adjusted 90-day odds ratio, 1.7 [CI, 1.5 to 2.1]; hazard ratios, 1.5 [CI, 1.2 to 1.8] Patients: All 3812 persons in Denmark who underwent splenec- from 91 to 365 days after splenectomy and 1.2 [CI, 1.1 to 1.4] tomy from 1996 to 2005. Splenectomized patients were matched beyond 365 days after splenectomy). Relative risks for infection to 3 comparison cohorts: the general population, appendec- were highest in patients who had splenectomy because of hema- tomized patients, and unsplenectomized patients with indica- tologic disorders. tions for splenectomy. Limitation: Increased surveillance among splenectomized patients Measurements: Relative risks were assessed for hospital contact may have affected the findings. involving any infection, pneumonia, and microbiologically confirmed bacteremia among 3812 splenectomized patients and their Conclusion: Splenectomy is associated with increased long-term matched comparisons, during different follow-up periods and after risk for infections involving hospital contact. regression analysis for confounder adjustment. Primary Funding Source: Amgen, Clinical Epidemiological Re- Results: The adjusted relative risk for any hospital contact with search Foundation at Aarhus University, and Karen Elise infection was highest within 90 days of splenectomy: 10.2% vs. Jensen Foundation. 0.6% among general population comparisons (adjusted odds ratio, 18.1 [95% CI, 14.8 to 22.1]) and 10.2% vs. 4.2% among appen- Ann Intern Med. 2009;151:546-555. www.annals.org dectomized patients (adjusted odds ratio, 2.4 [CI, 2.1 to 2.8]). The For author affiliations, see end of text.

he spleen hosts important immune cells that are essential for hospital contacts involving infection among splenecto- Tto the elimination of bloodborne pathogens, particularly mized patients compared with the general population and encapsulated bacteria (1, 2). Medically indicated splenectomy with similar groups of unsplenectomized patients. We also is associated with pneumonia, bacteremia, severe sepsis, and examined variations in the risk for infections by splenectomy other severe infections (3–9). The magnitude and duration of indication and by time since surgical procedure. increased infection risk, however, is debated (10). Few studies have followed splenectomized patients for more than 1 year (3, 5, 6, 8, 9), and confounding by splenectomy indications METHODS and other comorbid conditions could have affected the find- Cohort of Splenectomized Patients ings. In addition, the longer-term risk for infection in splenec- We identified all splenectomized patients from the tomized patients has not been quantified. Such data are im- Danish National Registry of Patients (NRP), which covers portant for evaluating splenectomy-associated risks and for the entire Danish population of 5.4 million persons. The better understanding patients’ clinical course (10). We con- NRP has tracked 99.4% of all discharges from Danish ducted a population-based assessment in Denmark of the risk acute care, nonpsychiatric hospitals since 1977 and all hospital outpatient and emergency department visits since 1995 (11). Records include dates of admission and dis- See also: charge; surgical procedure codes; and up to 20 discharge diagnoses, classified according to the International Classi- Print fication of Disease, Eighth Edition, until 31 December Editors’ Notes ...... 547 1993 and the Tenth Edition thereafter. In all Danish reg- Summary for Patients...... I-42 istries, patients are identified by their civil registration Web-Only number. These unique identifiers are stored in the Danish Appendix Civil Registration System (CRS) along with birth date; Conversion of graphics into slides residency status; and dates of immigration, emigration, and death (if applicable) (12).

From the NRP, we identified all surgical splenectomy procedures from 1 January 1996 to 31 December 2005. Context We also searched the NRP for patients with any previous Splenectomy is associated with infection, but the magni- medical indication for splenectomy, including diagnoses tude of risk is unclear. made during the hospitalization with splenectomy. We Contribution classified splenectomized patients into 8 groups according to indication. In cases of multiple indications, the category This study compared the risk for infections that required was determined by using the following hierarchy, regard- hospital care in all 3812 patients in Denmark who under- less of other indications (13): traumatic rupture of spleen, went splenectomy from 1996 to 2005 with that of comparison groups. Infection risk was highest in the 90 days immune thrombocytopenic purpura (ITP), other or unspe- after splenectomy. The risk for infection was 4.6 times cific thrombocytopenia, hematopoietic cancer, hereditary higher in splenectomized patients than in the general pop- hemolytic anemia, abdominal cancer, splenomegaly or ulation. The risk was only modestly higher than that of other splenic diseases only, and no diagnosis recorded (Ap- patients who had no splenectomy but did have the condi- pendix, available at www.annals.org). tions that can lead to splenectomy. General Population Comparison Cohort Implication From the CRS, we randomly chose 10 members of the general population, matched by age and sex, for each sple- Although splenectomized patients have a high risk for nectomized patient. The general population members had infection, some of this risk is due to underlying conditions to be alive without history of splenectomy as of the sple- and not to splenectomy alone. nectomy date of their matched patient (the index date). For the subset of splenectomized patients living in former —The Editors North Jutland County (about 500 000 inhabitants), for whom detailed data on bacteremia episodes were available pneumonia after the index date. We analyzed for bactere- (14, 15), we randomly chose 10 age- and sex-matched pop- mia episodes by using the subset of persons from North ulation comparisons living in the same county. Jutland County, whose population-based bacteremia re- Patient Comparison Cohorts search registry enables identification of all clinically signif- Patients undergoing splenectomy are exposed to sur- icant bacteremia episodes and all available blood culture gery and may have other diseases that increase the risk for samples (14, 15). infection or that involve immunosuppressive therapies. To Other Covariates separate potential effects of those conditions from effects of To control for confounding by conditions associated splenectomy on the risk for infection, we assembled 2 ad- with both splenectomy and risk for infection, we retrieved ditional comparison groups: patients who had appendec- data on several comorbid conditions, identified as condi- tomy (the appendectomized comparison cohort, for whom tions recorded in the NRP before each patient’s index date the index date was the date of appendectomy) and, for (Appendix, available at www.annals.org). splenectomized patients with a recorded indication for Statistical Analysis splenectomy, patients who had the same condition diag- We assessed the association between splenectomy nosed (for instance, myeloid leukemia) but without sple- for any indication and subsequent infection and for each nectomy (the matched-indication comparison cohort). For splenectomy indication versus matched comparisons in these comparisons, up to 5 matched patients without sple- the general population, appendectomy, and matched- nectomy on the index date were identified by using the indication comparison cohorts. CRS and NRP. In the matched-indication-cohort, we also We observed the patients from the index date until a matched on year of diagnosis of the medical condition. We hospital contact involving an outcome of interest (any in- then calculated the index date in each matched-indication fection, pneumonia, or nonpneumonia infection), death, patient by adding the time between diagnosis of the med- emigration, or 31 December 2006, whichever came first. ical condition and splenectomy in the splenectomized pa- Referent patients who underwent splenectomy during the tient to the date of the diagnosis. For patients who had follow-up were censored at the time of splenectomy. We splenectomy because of trauma, we selected trauma patient first compared overall rates of infectious events in the co- comparisons who underwent surgery for acute injury of horts during total follow-up. the spleen, liver, or gallbladder but not splenectomy (the We then compared occurrence of infection requiring Appendix, available at www.annals.org, provides the codes hospitalization in patients with and without splenectomy used). within 90 days, from 91 to 365 days, and more than 365 Hospital Contacts Involving Infection days after the index date. Because the NRP does not list the All participants’ records were linked to the CRS and exact date of the infectious event during the hospitalization the NRP to identify all hospital contacts involving bacterial in which the splenectomy was performed (4, 6), we esti- and viral infections, pneumonia, and infections other than mated odds ratios with 95% CIs, by using logistic regres- www.annals.org 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 547 Improving Patient Care Risk for Infection After Splenectomy

Table 1. Characteristics of Study Sample at Baseline

Characteristic Splenectomized Cohort General Population Appendectomized Patient Patients With n (%) Comparison Cohort Comparison Cohort Splenectomy Indication ,(3812 ؍ n) (%) n ,(8310 ؍ n (%) (n ,(962 16 ؍ n (%) (n ,(120 38 ؍ n) Age 0–39 y 916 (24.0) 9176 (24.1) 4580 (27.0) 2008 (24.2) 40–59 y 1015 (26.6) 10 112 (26.5) 5058 (29.8) 1774 (21.4) 60–69 y 751 (19.7) 7543 (19.8) 3245 (19.1) 1379 (16.6) Ն70 y 1130 (29.6) 11 289 (29.6) 4079 (24.1) 3149 (37.9)

Sex Female 1641 (43.1) 16 410 (43.1) 7973 (47.0) 3968 (47.8) Male 2171 (57.0) 21 710 (57.0) 8989 (53.0) 4342 (52.3)

Comorbid condition Myocardial infarction 167 (4.4) 1300 (3.4) 645 (3.8) 378 (4.6) Congestive heart failure 120 (3.2) 880 (2.3) 438 (2.6) 391 (4.7) Peripheral vascular disease 229 (6.0) 863 (2.3) 417 (2.5) 321 (3.9) Cerebrovascular disease 198 (5.2) 1762 (4.6) 800 (4.7) 508 (6.1) Chronic pulmonary disease 266 (7.0) 1744 (4.6) 859 (5.1) 587 (7.1) Connective tissue disease 117 (3.1) 652 (1.7) 331 (2.0) 298 (3.6) Inflammatory bowel disease 101 (2.7) 338 (0.9) 202 (1.2) 151 (1.8) Peptic ulcer disease 293 (7.7) 1060 (2.8) 593 (3.5) 642 (7.7) Liver disease 93 (2.4) 244 (0.6) 121 (0.7) 223 (2.7) Renal disease 100 (2.6) 286 (0.8) 157 (0.9) 158 (1.9) Diabetes 140 (3.7) 1090 (2.9) 479 (2.8) 406 (4.9) Hemiplegy 5 (0.1) 50 (0.1) 27 (0.2) 20 (0.2) Obesity 62 (1.6) 449 (1.2) 256 (1.5) 153 (1.8) Pancreatitis 157 (4.1) 220 (0.6) 134 (0.8) 128 (1.5) Alcoholism and alcoholism-related conditions 190 (5.0) 882 (2.3) 442 (2.6) 311 (3.7) Malignant conditions 1065 (28.0) 2227 (5.8) 1722 (10.2) 3327 (40.0) AIDS 6 (0.2) 8 (0.0) 5 (0.0) 11 (0.1) Dementia 20 (0.5) 206 (0.5) 81 (0.5) 59 (0.7) sion as the measure of relative risk for infections within 90 North Carolina). The Danish Data Protection Agency and days of the index date. We counted all discharge diagnoses Aarhus University Hospital Registry Board approved the of infection documented either for ongoing hospitaliza- study. tions ending within 90 days or for new hospital contacts Role of the Funding Source occurring within 90 days after the surgery date. This study was supported in part by a grant from For infections more than 90 days after splenectomy, Amgen to Aarhus University, by the Clinical Epidemiolog- we used Cox proportional hazards regression to compute ical Research Foundation at Aarhus University, and by the hazard ratios as measures of relative risk separately for the Ͼ Karen Elise Jensen Foundation. In collaboration with the intermediate (91 to 365 days) and long-term ( 365 days) investigators, Amgen designed the study. Amgen represen- follow-up. Thus, for each period we computed the time to tatives participated in the interpretation of the data, which the first infectious event among all patients alive and at risk Aarhus University holds, and in the writing of this report. at days 91 and 366, respectively, also including participants who had a hospital contact with infection during a previous period. We conducted similar analyses for bacteremia RESULTS in the North Jutland County subcohort (6, 16). Descriptive Data We adjusted for age (0 to 39, 40 to 59, 60 to 69, and We identified 3812 splenectomized persons, 38 120 Ն70 years), sex, and comorbid conditions (Appendix, general population members, 16 962 appendectomized co- available at www.annals.org) as a priori confounders. In the hort members, and 8310 matched-indication patients for matched-indication cohort analysis, presence of comorbid the 2394 splenectomized patients with recorded indica- conditions was dichotomized (yes or no) because the sam- tion. The median patient age was 60 years (interquartile ple was small. We used stratified regression models to ac- range, 41 to 72 years). Splenectomized patients and count for matching. The traumatic rupture comparisons matched-indication comparisons had a higher burden of were unmatched because of sparse data; therefore, the over- comorbid conditions than the general population or ap- all estimates of splenectomy versus indication comparisons pendectomized comparison group (Table 1). were calculated as an inverse variance weighted average of The most common medical indications for splenec- the individual indication estimates. We analyzed data by tomy were traumatic rupture of the spleen (20.1%) and using SAS software, version 9.2 (SAS Institute, Cary, abdominal cancer (19.3%). Other indications were spleno-

548 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 www.annals.org Risk for Infection After Splenectomy Improving Patient Care megaly (11.9%; often combined with other indications, Table 2. Indications Recorded Before or During Admission such as hematopoietic cancer), hematopoietic cancer for Splenectomy Among Splenectomized Patients (7.6%), and ITP (7.1%; often combined with nonspeciﬁc thrombocytopenia codes) (Table 2). Indication Splenectomized Cohort *(%) n ,(3812 ؍ n) Risk for Hospital Contacts With Infection Traumatic rupture of spleen 765 (20.1) Among the 3812 splenectomized patients, 955 Immune thrombocytopenic purpura 269 (7.1) (25.1%) had at least 1 hospital contact involving infection Other or nonspecified thrombocytopenia 210 (5.5) during a median follow-up of 2.2 years. The respective Hematopoietic cancer 288 (7.6) Hereditary hemolytic anemia 145 (3.8) summary measures were 10.8% and 5.1 years in the gen- Abdominal cancer 736 (19.3) eral population comparison cohort, 14.9% and 4.7 years in Splenomegaly or splenic disease 454 (11.9) the appendectomized comparison cohort, and 20.2% and None of these recorded 1418 (37.2)

2.1 years in the matched-indication comparison cohort. * Number of patients according to medical indications do not add up to 3812 The corresponding overall incidence rates of infection were because this table includes 429 patients (11.3% of splenectomized patients) who had more than 1 indication recorded. Frequent combinations were immune 7.7 per 100 person-years in splenectomized patients, 2.0 per thrombocytopenic purpura and nonspecified thrombocytopenia (157 patients 100 person-years in the general population cohort, 3.0 per [4.1%]) and hematopoietic cancer and splenomegaly or splenic disease (142 pa- 100 person-years in the appendectomized cohort, and 6.4 per tients [3.7%]). 100 person-years in the matched-indication comparison cohort. versus the appendectomized comparison cohort, and 3.2 The highest increase in the risk for infection was seen (CI, 1.7 to 6.1) versus the matched-indication compari- within the first 90 days, with 10.2% prevalence among sons. Adjusted hazard ratios were 13.5 (CI, 4.4 to 41.4) splenectomized patients versus 0.6% among the general and 1.9 (CI, 1.1 to 3.6) for bacteremia occurring from 91 population cohort (adjusted odds ratio, 18.1 [95% CI, to 365 days after splenectomy and after 365 days, respec- 14.8 to 22.1]); the increase in risk was similar for pneu- tively, compared with the general population, and these monia and nonpneumonia infections (Table 3). The 90- hazard ratios were lower than those in the appendecto- day estimates were lower but still increased compared with mized comparison cohort (Table 3). Too few episodes of those in appendectomized patients (adjusted odds ratio, bacteremia existed within splenectomy indication sub- 2.4 [CI, 2.1 to 2.8]) and matched-indication comparisons cohorts to calculate weighted overall longer-term hazard (adjusted odds ratio, 1.7 [CI, 1.5 to 2.1]). The estimates ratios. The abdomen was a much more frequent focus of changed little after we excluded 1086 patients with sple- bacteremia among splenectomized and appendectomized nectomy more than 2 days after the initial hospitalization; patients than among the other cohorts, whereas a urinary in these patients, infection could in theory have preceded tract focus was relatively rarer in the splenectomized and splenectomy (data not shown). splenectomy indication cohorts (Table 4). The distribution Compared with the general population, the hazard ra- of microbial agents was similar between groups. Of note, tio of infection among the splenectomized patients was 4.6 encapsulated bacteria, such as pneumococci, meningo- (CI, 3.8 to 5.5) from 91 to 365 days after splenectomy and cocci, and Haemophilus influenzae, were rarely encountered 2.5 (CI, 2.2 to 2.8) more than 365 days after splenectomy. in the splenectomized cohort. We observed a similar pattern and magnitude of effect with Laparoscopic splenectomy was performed in 33 pa- the appendectomized comparison cohort and lower, al- tients, 2 of whom had an infection involving hospital con- though still increased, estimates with the matched- tact within 90 days; another 5 had an infection more than indication comparison cohort (Table 3 and Figure). The 365 days after the procedure. This corresponds to overall risk estimates were similarly increased for pneumonia and infection incidence rates per 100 person-years of 5.5 nonpneumonia infections. The distribution of nonpneu- among those undergoing laparoscopic splenectomy and 7.6 monia infection types was similar between cohorts, with among those undergoing a nonlaparoscopic procedure. For septicemia, gastrointestinal infections, respiratory tract in- the period of more than 365 days after splenectomy, the fections, and miscellaneous or unspecified bacterial infec- risk for infections involving hospital contact among pa- tions representing the most frequent infections (data not tients who underwent laparoscopic splenectomy was 4.2 shown). Analyses restricted to hospital contacts after 90 times (CI, 1.3 to 14.0) higher than the corresponding risk days with infection as the first-listed diagnosis (Ͼ75% of for the general population. all contacts) did not substantially affect the relative risk Risk for Hospital Contact With Infection, by estimates (data not shown). Splenectomy Indication In the North Jutland County subcohort, 62 of 416 The adjusted odds of hospital contact with infection (14.9%) splenectomized patients had bacteremia (5.5 epi- 90 days after splenectomy were 14- to 46-fold higher in all sodes per 100 person-years) (Table 4). The adjusted 90- indication groups than in the general population compar- day odds ratios for bacteremia were 138.2 (CI, 41.5 to isons, whereas odds ratios ranged from 1.0 to 12.7 with 461.0) versus the general population, 3.6 (CI, 2.1 to 6.0) appendectomized patients. This pattern differed during www.annals.org 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 549 Improving Patient Care Risk for Infection After Splenectomy

Table 3. Risks for Hospital Contact Involving Infection and Microbiologically Confirmed Bacteremia, by Time Since Splenectomy

Time Since Splenectomy Splenectomized Infections Among Splenectomized Patients vs. Patients General Population Comparisons

Population Relative Risk (95% CI)* Comparisons Crude Adjusted 90-d follow-up, n (%) 3812 38 120 Any infection involving hospital contact 390 (10.2) 242 (0.6) 19.2 (16.1–22.8) 18.1 (14.8–22.1) Pneumonia 218 (5.7) 129 (0.3) 18.4 (14.7–23.1) 18.5 (14.2–24.2) Other infection† 206 (5.4) 126 (0.3) 18.4 (14.5–23.2) 17.8 (13.5–23.4) Microbiologically confirmed bacteremia‡ 41 (9.9) 3 (0.1) 151.5 (46.7–491.3) 138.2 (41.5–461.0)

91- to 365-d follow-up, n (n per 100 PYs) 3135 37 868 Any infection involving hospital contact 194 (9.0) 572 (2.0) 5.2 (4.3–6.2) 4.6 (3.8–5.5) Pneumonia 92 (4.2) 315 (1.1) 4.7 (3.7–6.1) 4.4 (3.3–5.9) Other infection† 130 (6.0) 292 (1.0) 6.3 (5.1–7.9) 5.5 (4.3–7.0) Microbiologically confirmed bacteremia‡ 11 (5.0) 6 (0.2) 24.5 (9.1–66.4) 13.5 (4.4–41.4)

>365-d follow-up, n (n per 100 PYs) 2787 37 087 Any infection involving hospital contact 528 (5.0) 3562 (2.1) 2.9 (2.6–3.2) 2.5 (2.2–2.8) Pneumonia 290 (2.6) 2169 (1.3) 2.7 (2.4–3.1) 2.3 (2.0–2.7) Other infection† 348 (3.2) 1871 (1.1) 3.4 (3.0–3.8) 2.9 (2.5–3.3) Microbiologically confirmed bacteremia‡ 14 (1.5) 88 (0.6) 2.6 (1.5–4.5) 1.9 (1.1–3.6)

NA ϭ not applicable; PY ϭ person-year. * Relative risk is estimated as the odds ratio of infection within 90 days and as the hazard ratio of infection within 91 to 365 days and more than 365 days after splenectomy. Adjusted estimates are adjusted by age, sex, and presence of comorbid condition at the time of splenectomy. † Infections other than pneumonia involving hospital contact. ‡ Microbiologically conﬁrmed bacteremia episodes assessed among 416 splenectomized patients (258 with a recorded indication) living in former North Jutland County and their residency-matched 4160 population comparisons, 1530 appendectomized comparisons, and 764 matched-indication comparisons. § Too few bacteremia events occurred within an individual splenectomy indication subcohorts to calculate a weighted overall hazard ratio for bacteremia. later follow-up periods (Table 5). Compared with the gen- the 91- to 365-day period and 5.8 and 5.2 after more than eral population and appendectomized comparison cohorts, 365 days of follow-up. By comparison, elderly patients the increase in the infection hazard among patients with who had splenectomy after abdominal cancer (80% were underlying hematopoietic cancer was 13.2 and 8.6 during aged Ն60 years) had only a 1.3- to 1.4-fold increased hazard of infection after 365 days of follow-up. Among relatively young and previously healthy persons who had Figure. Adjusted relative risk for any infection involving splenic trauma (82% aged Ͻ60 years), the hazard of infec- hospital contact among splenectomized patients compared tion remained 1.8- to 2.5-fold higher than in the appen- with general population, appendectomized, and dectomized or general population comparison cohorts matched-indication cohorts. more than 365 days after splenectomy. For patients with ITP, the corresponding hazard ratios were 2.6 to 4.0 (Table 5). We generally found that the odds of 90-day infection were increased 2- to 3-fold when comparing splenectomized patients with unsplenectomized patients matched on the same nontraumatic indication for splenectomy. In the long term, hazard ratios approached 1.0, except in patients with ITP (hazard ratio, 1.4 [CI, 1.0 to 2.0]) and hematopoietic cancer (hazard ratio, 1.6 [CI, 1.2 to 2.2]). Among patients splenectomized because of trauma, the 90-day risk for infection was not increased (odds ratio, 0.8 [CI, 0.5 to 1.2]) compared with matched trauma patients with abdominal surgery; however, the corresponding hazard ratio after 1 year was 1.3 (CI, 0.9 to 1.9).

Relative risk is estimated as the odds ratio of infection within 90 days DISCUSSION and as the hazard ratio of infection within 91 to 365 days and more than 365 days after splenectomy. Estimates are adjusted by age, sex, and pres- Our large nationwide, population-based study pro- ence of comorbid condition at the time of splenectomy. vides strong evidence of an increased risk for pneumonia

550 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 www.annals.org Risk for Infection After Splenectomy Improving Patient Care

Table 3—Continued

Infections Among Splenectomized Patients vs. Infections Among Splenectomized Patients With Recorded Indication vs. Appendectomized Comparisons Matched-Indication Comparisons

Appendectomized Relative Risk (95% CI)* Matched-Indication Relative Risk (95% CI)* Comparisons Comparisons Crude Adjusted Crude Adjusted 16 962 8310 713 (4.2) 2.5 (2.2–2.9) 2.4 (2.1–2.8) 479 (5.8) 1.7 (1.4–2.0) 1.7 (1.5–2.1) 346 (2.0) 2.7 (2.3–3.3) 2.7 (2.2–3.3) 230 (2.8) 2.0 (1.6–2.6) 2.2 (1.7–2.8) 403 (2.4) 2.3 (1.9–2.7) 2.1 (1.8–2.6) 279 (3.4) 1.4 (1.1–1.7) 1.4 (1.1–1.8) 42 (2.8) 3.7 (2.2–6.2) 3.6 (2.1–6.0) 25 (3.3) 2.5 (1.4–4.6) 3.2 (1.7–6.1)

16 469 7087 354 (2.9) 3.4 (2.8–4.1) 2.9 (2.3–3.6) 404 (8.6) 1.4 (1.1–1.7) 1.5 (1.2–1.8) 159 (1.3) 3.5 (2.6–4.6) 2.9 (2.1–4.1) 191 (4.0) 1.5 (1.0–2.1) 1.7 (1.2–2.5) 217 (1.8) 3.7 (2.9–4.7) 3.2 (2.5–4.2) 269 (5.7) 1.4 (1.1–1.8) 1.5 (1.1–1.9) 12 (1.2) 7.2 (2.4–21.6) 5.2 (1.7–16.3) 10 (2.7) NA§ NA§

15 942 5944 1722 (2.4) 2.2 (2.0–2.5) 2.0 (1.7–2.2) 1053 (5.0) 1.2 (1.0–1.3) 1.2 (1.1–1.4) 940 (1.3) 2.4 (2.0–2.8) 2.2 (1.8–2.6) 583 (2.6) 1.2 (1.0–1.5) 1.3 (1.1–1.7) 995 (1.4) 2.4 (2.0–2.7) 2.0 (1.8–2.4) 657 (3.0) 1.2 (1.0–1.5) 1.3 (1.1–1.5) 34 (0.6) 2.0 (0.9–4.4) 1.2 (0.5–2.8) 30 (2.1) NA§ NA§ and other infections, in particular early bacteremia, among serious infections per 100 person-years reported by patients undergoing splenectomy. This excess risk was Schwarz and colleagues (9) more than 30 years ago. In a most pronounced during the ﬁrst 90 days after splenec- recent Scottish cohort study of 1648 splenectomized pa- tomy, persisted for more than 365 days, and was seen re- tients, the risk for hospitalizations for infections was 21% gardless of the medical indication for splenectomy. with a mean follow-up of 4.5 years (excluding the ﬁrst 28 Our data are largely consistent with previous observa- days). As in our study, patients splenectomized because of tions of a high incidence of infections after splenectomy. hematologic malignant conditions had the highest rates of We observed an incidence similar to the estimate of 7.2 infection, and patients with splenic trauma had the lowest

Table 4. Bacteremia-Related Characteristics in Splenectomized Patients and Comparisons

Characteristic Splenectomized Cohort General Population Appendectomized Patient Splenectomy Indication Comparison Comparison Comparison (416 ؍ n) (764 ؍ Cohort (n (1530 ؍ Cohort (n (4160 ؍ Cohort (n Microbiologically confirmed bacteremia during total 62 (5.5) 97 (0.5) 87 (1.3) 62 (3.3) follow-up, n (n per 100 PYs)

Focus of infection, n (%)* Urinary tract 8 (13) 16 (31) 22 (25) 10 (16) Abdominal 24 (39) 3 (6) 25 (29) 13 (21) Respiratory tract 4 (6) 4 (8) 3 (3) 8 (13) Skin, central nervous system, or other 13 (20) 15 (29) 20 (22) 15 (24) Undetermined focus 13 (21) 14 (27) 17 (20) 16 (26)

Microbial agent group, n (%)* Streptococcus pneumoniae 2 (3) 3 (6) 2 (2) 5 (8) Staphylococcus aureus 10 (16) 6 (12) 6 (7) 7 (11) Enterococci 4 (6) 3 (6) 0 (0) 0 (0) Other gram-positive bacteria 2 (3) 3 (6) 13 (15) 5 (8) Escherichia coli 14 (23) 15 (29) 31 (36) 21 (34) Other enterobacteria 10 (16) 8 (15) 7 (8) 7 (11) Other gram-negative rods (including Haemophilus 0 (0) 3 (6) 2 (2) 1 (2) influenzae and meningococci) Pseudomonas aeruginosa 5 (8) 3 (6) 2 (2) 2 (3) Other microorganisms (anaerobic bacteria, 15 (24) 8 (15) 24 (28) 14 (23) Candida species, polymicrobial growth)

PY ϭ person-year. * Percentages are based on total of all infectious foci and microbial agent groups. www.annals.org 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 551 Improving Patient Care Risk for Infection After Splenectomy

Table 5. Risks for Hospital Contact Involving Infection in Splenectomized Patients, by Indication and Time Since Splenectomy*

Indication for Splenectomy Splenectomized Infections Among Splenectomized Infections Among Splenectomized Infections Among Patients Patients vs. General Population Patients vs. Appendectomized Splenectomized Patients Comparisons Comparisons vs. Matched-Indication Comparisons

Population Adjusted Relative Appendectomized Adjusted Matched- Adjusted Comparisons Risk (95% CI)† Comparisons Relative Risk Indication Relative Risk (95% CI)† Comparisons (95% CI)† Traumatic rupture of spleen Patients, n 765 7650 3674 516 Median age (interquartile range), 36 (21–54) 36 (21–54) 35 (22–52) 33 (21–51) y Men, % 73.5 73.5 72.6 65.5 Persons with comorbid condition, 24.7 15.7 18.9 26.2 % Median days since splenectomy 0– – 0 indication Infection involving hospital contact 0to90d,n (%) 59 (7.7) 44 (0.6) 21.2 (12.8–35.1) 127 (3.5) 2.0 (1.4–2.9) 51 (9.9) 0.8 (0.5–1.2) 91 to 365 d, n (n per 100 PYs) 23 (4.6) 70 (1.2) 3.4 (2.0–5.9) 56 (2.1) 2.9 (1.6–5.2) 10 (3.3) 1.4 (0.6–2.9) Ͼ365 d, n (n per 100 PYs) 92 (3.3) 421 (1.2) 2.5 (2.0–3.3) 260 (1.6) 1.8 (1.3–2.3) 40 (2.3) 1.3 (0.9–1.9)

Immune thrombocytopenic purpura Patients, n 269 2690 1292 1345 Median age (interquartile range), 47 (28–61) 47 (28–61) 45 (28–59) 38 (11–66) y Men, % 36.4 36.4 34.2 43.6 Persons with comorbid condition, 27.9 16.7 23.3 33.1 % Median days since splenectomy 196 – – 196 indication Infection involving hospital contact 0to90d,n (%) 15 (5.6) 14 (0.5) 14.8 (5.1–43.2) 45 (3.5) 1.3 (0.6–2.5) 36 (2.7) 2.6 (1.3–5.1) 91 to 365 d, n (n per 100 PYs) 11 (5.7) 20 (1.0) 14.2 (4.3–46.1) 20 (2.1) 4.8 (1.7–13.6) 59 (6.5) 1.0 (0.5–2.0) Ͼ365 d, n (n per 100 PYs) 56 (4.6) 176 (1.3) 4.0 (2.8–5.6) 106 (1.7) 2.6 (1.8–3.8) 173 (3.3) 1.4 (1.0–2.0)

Other or nonspecified thrombocytopenia Patients, n 53 530 262 265 Median age (interquartile range), 56 (40–67) 56 (40–67) 55 (41–68) 52 (30–66) y Men, % 50.9 50.9 50.4 52.8 Persons with comorbid condition, 64.2 21.9 27.9 50.9 % Median days since splenectomy 204 – – 204 indication Infection involving hospital contact 0to90d,n (%) 15 (28.3) 2 (0.4) 46.6 (5.4–405.9) 10 (3.8) 4.5 (1.3–16.2) 25 (9.4) 3.5 (1.5–7.9) 91 to 365 d, n (n per 100 PYs) 11 (36.8) 5 (1.3) 94.6 (10.0–895.7) 6 (3.1) 25.0 (3.6–174.2) 30 (18.2) 2.2 (1.0–4.7) Ͼ365 d, n (n per 100 PYs) 13 (9.6) 49 (2.1) 4.8 (2.0–12.0) 26 (2.5) 2.0 (0.8–5.4) 43 (4.8) 1.3 (0.6–2.9)

Hematopoietic cancer Patients, n 252 2520 1092 1252 Median age (interquartile range), 65 (54–72) 65 (54–71) 62 (52–70) 65 (51–76) y Men, % 56.0 56.0 50.3 55.2 Persons with comorbid condition, 83.7 28.4 35.9 89.2 % Median days since splenectomy 297 – – 296 indication Infection involving hospital contact 0to90d,n (%) 40 (15.9) 21 (0.8) 25.6 (6.3–104.8) 50 (4.6) 3.5 (1.6–7.5) 98 (7.8) 2.4 (1.6–3.6) 91 to 365 d, n (n per 100 PYs) 37 (27.1) 42 (2.3) 13.2 (5.5–31.5) 26 (3.3) 8.6 (3.2–22.7) 110 (14.2) 2.1 (1.4–3.2) Ͼ365 d, n (n per 100 PYs) 69 (14.1) 246 (2.3) 5.8 (3.8–8.8) 122 (2.7) 5.2 (3.2–8.7) 285 (8.8) 1.6 (1.2–2.2)

Hereditary hemolytic anemia Patients, n 142 1420 705 618 Median age (interquartile range), 14 (8–32) 14 (9–32) 14 (9–31) 27 (9–49) y

Continued on following page

552 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 www.annals.org Risk for Infection After Splenectomy Improving Patient Care

Table 5—Continued

Population Adjusted Relative Appendectomized Adjusted Matched- Adjusted Comparisons Risk (95% CI)† Comparisons Relative Risk Indication Relative Risk (95% CI)† Comparisons (95% CI)† Men, % 47.9 47.9 47.5 47.7 Persons with comorbid condition, 12.7 8.5 12.6 26.1 % Median days since splenectomy 1874 – – 1911 indication Infection involving hospital contact 0to90d,n (%) 8 (5.6) 5 (0.4) 31.5 (6.4–153.7) 36 (5.1) 1.0 (0.4–2.5) 11 (1.8) 3.6 (1.1–12.5) 91 to 365 d, n (n per 100 PYs) 6 (5.8) 13 (1.2) 4.3 (1.4–13.4) 8 (1.5) 7.8 (1.8–34.1) 27 (6.4) 1.0 (0.4–2.6) Ͼ365 d, n (n per 100 PYs) 18 (3.1) 61 (1.0) 4.0 (2.3–7.2) 45 (1.5) 2.3 (1.2–4.2) 79 (3.7) 1.1 (0.6–1.9)

Abdominal cancer Patients, n 712 7120 2857 3560 Median age (interquartile range), 72 (62–79) 72 (62–79) 70 (60–78) 72 (63–80) y Men, % 54.2 54.2 46.0 53.0 Persons with comorbid condition, 70.7 35.4 43.9 73.9 % Median days since splenectomy 11 – – 11 indication Infection involving hospital contact 0to90d,n (%) 59 (8.3) 60 (0.8) 15.3 (9.2–25.6) 131 (4.6) 1.9 (1.3–2.9) 155 (4.4) 2.0 (1.5–2.8) 91 to 365 d, n (n per 100 PYs) 33 (9.4) 144 (2.8) 3.6 (2.3–5.7) 77 (3.9) 2.5 (1.4–4.5) 107 (6.3) 1.4 (0.9–2.2) Ͼ365 d, n (n per 100 PYs) 65 (5.4) 966 (3.4) 1.4 (1.1–2.0) 391 (3.7) 1.3 (0.9–1.9) 308 (5.2) 1.0 (0.7–1.4)

Splenomegaly or splenic disease Patients, n 201 2010 942 754 Median age (interquartile range), 46 (33–63) 46 (33–63) 45 (31–61) 46 (33–63) y Men, % 57.7 57.7 54.9 53.6 Persons with comorbid condition, 46.8 19.6 25.6 50.0 % Median days since splenectomy 2– – 3 indication Infection involving hospital contact 0to90d,n (%) 36 (17.9) 4 (0.2) 118.2‡ (36.4–384.1) 24 (2.6) 12.7 (6.0–27.1) 103 (13.7) 1.5 (0.9–2.4) 91 to 365 d, n (n per 100 PYs) 11 (8.3) 23 (1.5) 17.0 (4.7–61.4) 16 (2.3) 10.0 (2.0–50.4) 61 (13.7) 0.6 (0.3–1.3) Ͼ365 d, n (n per 100 PYs) 34 (5.4) 116 (1.4) 3.6 (2.2–5.9) 76 (2.0) 2.3 (1.3–4.1) 125 (6.7) 0.7 (0.4–1.2)

No specified indication for splenectomy recorded Patients, n 1418 14 180 6138 NA NA Median age (interquartile range), 65 (53–74) 65 (53–74) 62 (51–72) y Men, % 54.5 54.5 49.4 Persons with comorbid condition, 67.5 29.1 35.7 % Infection involving hospital contact 0to90d,n (%) 158 (11.1) 92 (0.7) 18.3 (13.3–25.1) 290 (4.7) 2.3 (1.8–2.95) 91 to 365 d, n (n per 100 PYs) 62 (8.7) 255 (2.4) 3.7 (2.7–5.2) 145 (3.4) 2.4 (1.6–3.6) Ͼ365 d, n (n per 100 PYs) 181 (5.7) 1527 (2.5) 2.1 (1.7–2.5) 696 (2.8) 1.8 (1.4–2.2)

NA ϭ not applicable (matched-indication comparisons applicable for only splenectomized patients with a speciﬁed splenectomy indication recorded); PY ϭ patient-year. * Includes patients within mutually exclusive categories of splenectomy indications; see text for classiﬁcation. † Relative risk estimated as the odds ratio of infection within 90 days and the hazard ratio of infection within 91 to 365 days and Ͼ365 days after splenectomy. Estimates adjusted by age, sex, and presence of any comorbid condition at the time of splenectomy. ‡ Crude 90-day odds ratio shown; too few events occurred to calculate adjusted 90-day odds ratio.

www.annals.org 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 553 Improving Patient Care Risk for Infection After Splenectomy rates. However, we also found high rates of infection for pneumonia (19) and close to 100% for bacteremia among patients splenectomized because of benign hemato- (20). The surgical procedure data we used are also known logic disorders. A previous cohort study of 98 patients sple- to have high validity (21), and any misclassification is likely nectomized for trauma and 98 control participants under- to attenuate our relative risk estimates. going laparotomy without splenectomy reported more Another concern is that increased surveillance of sple- early postoperative infections in the splenectomized cohort nectomized patients may have led to an overestimation of (45% vs. 30%; P ϭ 0.040), most of which were pneumo- infection risk. Prophylactic guidelines for splenectomized nia (30% vs. 14%; P ϭ 0.020) (4). Several studies have patients in Denmark include information about the risk reported a high risk for bacteremia and severe sepsis very for serious infections, and instructions to take a dose of early after the splenectomy procedure (3, 5–7, 9). Ejstrud penicillin V at home and to seek medical care in case of and colleagues (6) found a risk for bacteremia of 7% dur- pyrexia greater than 38.5 °C (22, 23). However, for such ing 3.2 years of follow-up, with 45% of episodes occurring severe infections as bacteremia, which had the highest risk within 30 days after splenectomy. The lack of comparison increase associated with splenectomy, the number of unde- groups or long-term follow-up in many earlier studies lim- tected cases in all comparison cohorts used is probably low ited their potential for clarifying the clinical impact of sple- because of severe symptoms and need for hospitalization. nectomy on infections. Although splenectomized patients may also be more likely The mechanisms underlying the association between to experience nosocomial infections because of more fre- splenectomy and infections are not entirely clear. Although quent hospital contacts, analyses restricted to hospital con- splenectomy and infections may share risk factors, such as tacts in which infection constituted the primary diagnosis cancer and chemotherapy, we also found markedly in- yielded virtually unchanged estimates. creased long-term rates of infection among patients under- In Denmark, pneumococcal vaccination is recom- going splenectomy because of disorders other than cancer. mended within 2 weeks before elective splenectomy, or as Modest associations persisted for most splenectomy indica- soon as possible and within less than 2 weeks after emer- tions even after we used appendectomized or matched- gent splenectomy (22–25). However, previous studies in indication comparison groups. In particular, encapsulated Denmark and elsewhere found that between 10% and bacteria (such as Streptococcus pneumoniae) have been asso- 40% of splenectomized patients received no pneumococcal ciated with early severe postsplenectomy infections. How- vaccination, and an even larger percentage was not vacci- ever, our study corroborates earlier findings (7) in which nated at the recommended time (6, 23, 26–29). Unfortu- pneumococci caused only 4% of bacteremia episodes. In nately, we did not have data on vaccinations, yet higher contrast to common belief, enteric rods seem to be the vaccination rates in patients with splenectomy probably predominant cause of early and late postsplenectomy would have caused underestimation of their relative risks bacteremia. for bacteremia and pneumonia. Neither Haemophilus influ- Abdominal surgery itself may increase infection risk, as enzae type B vaccination nor continuous antibiotic prophy- also suggested by the less markedly increased early risks we laxis is currently recommended in Denmark. found compared with those in patients who underwent Our findings may have clinical implications. Within 3 appendectomy. Laparoscopic splenectomy may have months, bacteremia is diagnosed in almost 10% of splenec- lower infection rates than open surgery (17), although tomized patients, and more than 5% have hospital- in our cohort, we also observed infections requiring hos- diagnosed pneumonia. Thus, preventive measures should pitalization in the few patients undergoing laparoscopic be intensified (24). This includes pneumococcal vaccina- procedures. tion, which has been associated with decreased bacteremia Our study has several strengths, including examination risk in splenectomized persons (6). Use of continuous an- of hospital contacts with important clinical infections, such tibiotic prophylaxis remains controversial because of con- as pneumonia and bacteremia. In addition, our estimates cerns regarding patient adherence and bacterial resistance, are derived from a population-based cohort study with as well as lack of clear evidence for efficacy (28). negligible referral and diagnostic biases. The large popula- Our study showed that splenectomy is associated with tion we studied was well-defined, and follow-up for infec- increased short- and long-term risk for infections involving tions requiring hospitalization and death was complete. hospital contact. Although relative risks decrease after the We had access to long-term hospital registry data and com- initial months, they remain clearly elevated compared with plete outpatient clinic data. those in the general population or appendectomized pa- Despite the advantages of the high-quality data, our tients and modestly elevated compared with those in pa- database study has limitations. First, we had access to lim- tients with the same medical indications for splenectomy. ited clinical information. Second, the validity of our findings depends on the accuracy of coding for splenectomy From Aarhus University Hospital, Aalborg, Denmark, and Amgen, Ux- and infection end points. In this context, the predictive bridge, Middlesex, United Kingdom, and Thousand Oaks, California. value of hospital registry diagnoses of major infections is reportedly high (18) and, in our research registries, is 90% Grant Support: By Amgen, the Clinical Epidemiological Research Foun-

554 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 www.annals.org Risk for Infection After Splenectomy Improving Patient Care dation at Aarhus University, and the Karen Elise Jensens Foundation. 12. Frank L. Epidemiology. When an entire country is a cohort. Science. 2000; 287:2398-9. [PMID: 10766613] Potential Conflicts of Interest: Employment: W.M. Schoonen (Amgen), 13. Mellemkjoer L, Olsen JH, Linet MS, Gridley G, McLaughlin JK. Cancer J.P. Fryzek (Amgen). Stock ownership or options (other than mutual funds): risk after splenectomy. Cancer. 1995;75:577-83. [PMID: 7812926] W.M. Schoonen (Amgen), J.P. Fryzek (Amgen). 14. Søgaard M, Schønheyder HC, Riis A, Sørensen HT, Nørgaard M. Short- term mortality in relation to age and comorbidity in older adults with community-acquired bacteremia: a population-based cohort study. J Am Geriatr Reproducible Research Statement: Study protocol and statistical code: Soc. 2008;56:1593-600. [PMID: 18691276] Available from Dr. Sørensen (e-mail, [email protected]). Data set: The Dan- 15. Thomsen RW, Hundborg HH, Johnsen SP, Pedersen L, Sørensen HT, ish Data Protection Agency does not allow transferral of the project data Schønheyder HC, et al. Statin use and mortality within 180 days after bactere- to external research institutions. Most of the source data are stored with mia: a population-based cohort study. Crit Care Med. 2006;34:1080-6. [PMID: Statistics Denmark and can be made available for analysis from www.dst 16484926] .dk/research. 16. Kornum JB, Thomsen RW, Riis A, Lervang HH, Schønheyder HC, Sørensen HT. Type 2 diabetes and pneumonia outcomes: a population-based Requests for Single Reprints: Reimar W. Thomsen, MD, PhD, De- cohort study. Diabetes Care. 2007;30:2251-7. [PMID: 17595354] partment of Clinical Epidemiology, Aarhus University Hospital, Aalborg 17. Sampath S, Meneghetti AT, MacFarlane JK, Nguyen NH, Benny WB, Hospital Science and Innovation Center, Søndre Skovvej 15, DK-9000 Panton ON. An 18-year review of open and laparoscopic splenectomy for idio- pathic thrombocytopenic purpura. Am J Surg. 2007;193:580-3; discussion Aalborg, Denmark; e-mail, [email protected]. 583-4. [PMID: 17434359] 18. Schneeweiss S, Robicsek A, Scranton R, Zuckerman D, Solomon DH. Current author addresses and author contributions are available at www Veteran’s affairs hospital discharge databases coded serious bacterial infections .annals.org. accurately. J Clin Epidemiol. 2007;60:397-409. [PMID: 17346615] 19. Thomsen RW, Riis A, Nørgaard M, Jacobsen J, Christensen S, McDonald CJ, et al. Rising incidence and persistently high mortality of hospitalized pneu- References monia: a 10-year population-based study in Denmark. J Intern Med. 2006;259: 410-7. [PMID: 16594909] 1. Chaplin DD. Lymphoid tissues and organs. In: Paul WE, ed. Fundamental 20. Schønheyder HC, Søgaard M. [Hospital-acquired bacteraemia and fungae- Immunology. Philadelphia: Lippincott Williams & Wilkins; 2003:419-55. mia. A regional study with national implications]. Ugeskr Laeger. 2007;169: 2. Gordon S. Macrophages and the immune response. In: Paul WE, ed. Funda- 4175-9. [PMID: 18211787] mental Immunology. Philadelphia: Lippincott Williams & Wilkins; 2003:482- 21. Sorensen HT, Sabroe S, Olsen J. A framework for evaluation of secondary 97. data sources for epidemiological research. Int J Epidemiol. 1996;25:435-42. 3. Kyaw MH, Holmes EM, Toolis F, Wayne B, Chalmers J, Jones IG, et al. [PMID: 9119571] Evaluation of severe infection and survival after splenectomy. Am J Med. 2006; 22. Birgens HS, Bergmann OJ, Borregaard N, Hansen NE, Hasselbalch H, 119:276.e1-7. [PMID: 16490477] 4. Wiseman J, Brown CV, Weng J, Salim A, Rhee P, Demetriades D. Splenec- Hippe E, et al. Benigne hæmatologiske sygdomme hos voksne. Klaringsrapport. tomy for trauma increases the rate of early postoperative infections. Am Surg. Copenhagen: Danish Medical Assoc; 2003. 2006;72:947-50. [PMID: 17058742] 23. Ejstrud P, Hansen JB, Andreasen DA. Prophylaxis against pneumococcal 5. Bisharat N, Omari H, Lavi I, Raz R. Risk of infection and death among infection after splenectomy: a challenge for hospitals and primary care. Eur J post-splenectomy patients. J Infect. 2001;43:182-6. [PMID: 11798256] Surg. 1997;163:733-8. [PMID: 9373223] 6. Ejstrud P, Kristensen B, Hansen JB, Madsen KM, Schønheyder HC, 24. Newland A, Provan D, Myint S. Preventing severe infection after splenec- Sørensen HT. Risk and patterns of bacteraemia after splenectomy: a population- tomy [Editorial]. BMJ. 2005;331:417-8. [PMID: 16110056] based study. Scand J Infect Dis. 2000;32:521-5. [PMID: 11055658] 25. Statens Serum Institute. EPI-NEWS: National surveillance of communica- 7. Deodhar HA, Marshall RJ, Barnes JN. Increased risk of sepsis after splenec- ble diseases. Accessed at www.ssi.dk/sw3024.asp on 20 August 2009. tomy. BMJ. 1993;307:1408-9. [PMID: 8274896] 26. Ernstro¨m F, Miskowiak J. [Follow-up of patients after splenectomy at a local 8. Cullingford GL, Watkins DN, Watts AD, Mallon DF. Severe late postsple- hospital]. Ugeskr Laeger. 1998;160:2887-9. [PMID: 9599568] nectomy infection. Br J Surg. 1991;78:716-21. [PMID: 2070242] 27. Kyaw MH, Holmes EM, Chalmers J, Jones IG, Campbell H. A survey of 9. Schwartz PE, Sterioff S, Mucha P, Melton LJ 3rd, Offord KP. Postsplenec- vaccine coverage and antibiotic prophylaxis in splenectomised patients in Scot- tomy sepsis and mortality in adults. JAMA. 1982;248:2279-83. [PMID: land. J Clin Pathol. 2002;55:472-4. [PMID: 12037033] 7131680] 28. Waghorn DJ. Overwhelming infection in asplenic patients: current best prac- 10. Cadili A, de Gara C. Complications of splenectomy. Am J Med. 2008;121: tice preventive measures are not being followed. J Clin Pathol. 2001;54:214-8. 371-5. [PMID: 18456028] [PMID: 11253134] 11. Andersen TF, Madsen M, Jørgensen J, Mellemkjoer L, Olsen JH. The 29. Kinnersley P, Wilkinson CE, Srinivasan J. Pneumococcal vaccination after Danish National Hospital Register. A valuable source of data for modern health splenectomy: survey of hospital and primary care records. BMJ. 1993;307: sciences. Dan Med Bull. 1999;46:263-8. [PMID: 10421985] 1398-9. [PMID: 8274893]

www.annals.org 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 555 Annals of Internal Medicine

Current Author Addresses: Dr. Thomsen: Department of Clinical Ep- Injury of liver or gallbladder: S36.1 (ICD-10); 864.9 (ICD-8). idemiology, Aarhus University Hospital, Aalborg Hospital Science and Together with any of: Innovation Centre, Søndre Skovvej 15, DK-9000 Aalborg, Denmark. Explorative laparotomy: 40220, JAH00. Dr. Schoonen: Amgen, 1 Uxbridge Business Park, Sanderson Road, Ux- Laparoscopy: 40240, JAH01. bridge, Middlesex UB8 1DH, United Kingdom. Ms. Farkas, Mr. Riis, Mr. Jacobsen, and Dr. Sørensen: Department of Repair (including suture) of spleen: 48930, JMB10. Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Alle Operations on liver or biliary tract: 47000–48299, JJ, JK. 43-45, DK-8200 Århus Nord, Denmark. Dr. Fryzek: Amgen, One Amgen Center Drive, Thousand Oaks, CA Appendectomy 91360. Open appendectomy (excluding laparoscopic or with drain- age): 43000, JEA00. Author Contributions: Conception and design: D.K. Farkas, J. Jacob- sen, R.W. Thomsen, W.M. Schoonen, A. Riis, J.P. Fryzek, H.T. Comorbid Conditions Sørensen. Myocardial infarction: I21; I22; I23 (ICD-10); 410 (ICD-8). Analysis and interpretation of the data: R.W. Thomsen, W.M. Schoo- Congestive heart failure: I50; I11.0; I13.0; I13.2 (ICD-10); nen, D.K. Farkas, A. Riis, J. Jacobsen, J.P. Fryzek, H.T. Sørensen. 427.09; 427.10; 427.11; 427.19; 428.99; 782.49 (ICD-8). Drafting of the article: R.W. Thomsen, W.M. Schoonen, J.P. Fryzek. Vascular disease: I70; I71; I72; I73; I74; I77 (ICD-10); 440; Critical revision of the article for important intellectual content: D.K. 441; 442; 443; 444; 445 (ICD-8). Farkas, J. Jacobsen, R.W. Thomsen, W.M. Schoonen, J.P. Fryzek, H.T. Sørensen. Cerebrovascular disease: I60–I69; G45; G46 (ICD-10); Final approval of the article: R.W. Thomsen, W.M. Schoonen, D.K. 430–438 (ICD-8). Farkas, A. Riis, J. Jacobsen, J.P. Fryzek, H.T. Sørensen. Chronic pulmonary disease: J40–J47; J60–J67; J68.4; Administrative, technical, or logistic support: D.K. Farkas, J. Jacobsen. J70.1; J70.3; J84.1; J92.0; J96.1; J98.2; J98.3 (ICD-10); 490– Statistical expertise: W.M. Schoonen, D.K. Farkas, A. Riis, J. Jacobsen, 493; 515–518 (ICD-8). J.P. Fryzek. Connective tissue disease: M05; M06; M08; M09; M30– Collection and assembly of data: A. Riis. 36; D86 (ICD-10); 712; 716; 734; 446; 135.99 (ICD-8). Inflammatory bowel disease: K50–K52 (ICD-10); 563 (ICD-8). APPENDIX:SURGERY AND DIAGNOSIS CODES Peptic ulcer disease: K22.1; K25–K28 (ICD-10); 530.91; Diagnosis codes were assigned from the International Clas- 530.98; 531–534 (ICD-8). sification of Diseases (ICD), Eighth Edition or Tenth Edition. Liver disease: B15.0; B16.0; B16.2; B18; B19.0; K70–K74; Surgical procedure codes were assigned from the Danish Na- K76.0; K76.6; I85 (ICD-10); 070.00; 070.02; 070.04; 070.06; tional Board of Health Classification of Surgical Procedures until 070.08; 571; 573.00; 573.01; 573.04; 456.00–456.09 (ICD-8). 31 December 1995 and the Nordic Medico-Statistical Commit- Renal disease: I12; I13; N00–N05; N07; N11; N14; N17– tee (NOMESCO) Classification of Surgical Procedures (NCSP) N19; Q61 (ICD-10); 403; 404; 580–583; 584; 590.09; 593.19; thereafter (30). 753.10–753.19; 792 (ICD-8). Splenectomy Diabetes: E10-E11 (ICD-10); 249; 250 (ICD-8). Partial or total (transabdominal, laparoscopic, or transtho- Hemiplegy: G81; G82 (ICD-10); 344 (ICD-8). racic) splenectomy: 48900–48910, JMA. Obesity: E65; E66 (ICD-10); 277.99 (ICD-8). Pancreatitis: K85; K86.0; K86.1 (ICD-10); 577.00–577.19 Splenectomy Indications (ICD-8). Traumatic rupture of spleen: S36.0 (ICD-10); 865.9 (ICD-8). Alcoholism and alcoholism-related conditions: F10; G31.2; Hematopoietic cancer: C81–C96 (ICD-10); 200-207, G62.1; G72.1; I 42.6 K29.2; R78.0; T51; Z72.1 (ICD-10); 291; 275.59 (ICD-8). 303; 979; 980 (ICD-8). Hereditary hemolytic anemias: D55–D58 (ICD-10); 282 Malignant conditions: C00–C96 (ICD-10); 140–207 (ICD-8). (ICD-8). ITP: D69.3 (ICD-10); 287.10 (ICD-8). AIDS: B21–B24 (ICD-10); 079.83 (ICD-8). Other or nonspecific thrombocytopenia: D69.4, D69.5, Dementia: F00–F03; F05.1; G30 (ICD-10); 290.09– D69.6 (ICD-10); 287.11, 287.18, 287.19, 287.29 (ICD-8). 290.19; 293.09 (ICD-8). Splenic diseases and splenomegaly: D73, R16.1, R16.2 (ICD-10); 289.4, 782.89 (ICD-8). Infectious Outcomes Abdominal cancers: C16, C18 (ICD-10); 151, 153 (ICD-8). Bacterial and viral infections: A00–B99 (ICD-10). Other or unknown indications: Various codes (ICD-10); Upper and lower respiratory tract infections: J00–J22 various codes (ICD-8). (ICD-10). Matched Indication for Splenectomy Due to Trauma 30. Nordic Medico-Statistical Committee (NOMESCO). NCSP: The Any of: NOMESCO Classification of Surgical Procedures. Accessed at www.nordclass.uu Injury of spleen: S36.0 (ICD-10); 865.9 (ICD-8). .se/verksam/ncspe.htm on 24 August 2009.

W-174 20 October 2009 Annals of Internal Medicine Volume 151 • Number 8 www.annals.org