- Email: [email protected]
Relationship between death and infections among patients hospitalized in internal medicine departments: A prevalence and validation study
American Journal of Infection Control 42 (2014) 506-10
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
Relationship between death and infections among patients hospitalized in internal medicine departments: A prevalence and validation study Anne B. Boisen MB a, Michael Dalager-Pedersen MD a, b, Mette Søgaard DVM, PhD a, Rikke Mortensen MSc a, Reimar W. Thomsen MD, PhD a, * a b
Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Aarhus, Denmark Department of Infectious Diseases, Aalborg University Hospital, Aalborg, Denmark
Key Words: Bacterial infections and mycoses Infectious disease medicine Cause of death/vital statistics Epidemiology
Background: We assessed the burden of infection-related death at internal medicine departments in Denmark and the validity of 2 population-based registries for diagnoses of infection. Methods: We reviewed medical records of all inpatients who died at a large university hospital during 2008 with an infection diagnosis in the Cause of Death Registry (CDR) or Hospital Discharge Registry (HDR). We computed the positive predictive value of infection diagnoses and completeness of each registry with 95% confidence intervals (CIs) and the prevalence of infection-related deaths by capturerecapture analysis. Results: Among 458 deaths, 193 patients (42.1%) had an infection diagnosis of which 40.0% (183 out of 458) were verified. The positive predictive value of an infection diagnosis was 96.0% (95% CI, 92.0%98.3%) in the CDR and 95.3% (CI, 90.9%-97.9%) in the HDR. Completeness of the CDR was 79.2% (CI, 72.9%84.6%) and completeness of the HDR was 77.0% (CI, 70.6%-82.7%). By captureerecapture analysis we estimated that 43.2% of all deaths (198 out of 458) were related to infection. Conclusions: The large proportion of deaths in internal medicine departments that are related to infection emphasizes the need for optimized infection prevention and treatment strategies. CDR and HDR are valid sources for identifying infection and may supplement each other for increased completeness of infection-related death in epidemiologic research. Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Infectious diseases are among the most frequent causes of acute hospital admission both in low- and high-income countries.1-4 During past decades, hospitalization rates for severe infections such as pneumonia or bacteremia/sepsis have increased in the Western world.5-8 Moreover, approximately 10% of patients acutely admitted to a hospital acquire 1 infections during their hospital stay.9 The associated in-hospital mortality approaches 4% following hospitalization with any infection1 and 10%-15% for patients with pneumonia or bacteremia.5,6,10 As a result, a considerable proportion of all deaths are expected to be related to infections. In highincome countries, cancer and heart diseases are the most
* Address correspondence to Reimar W. Thomsen, MD, PhD, Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Olof Palmes Allé 43-45, DK-8200 Aarhus N, Denmark. E-mail address: [email protected] (R.W. Thomsen). Funding for this investigation was provided by Aalborg University Hospital. Conflicts of interest: None to report.
common underlying causes of death.11 According to data from the Danish Cause of Death Registry (CDR), cancer accounted for 30.4% and heart disease for 16.9% of all deaths in Denmark during 2011. In contrast, an infectious disease as coded in ICD-10 was the underlying cause in only 1.6% of deaths.12 Notwithstanding, a recent population-based study using Hospital Discharge Registry (HDR) data from northern Denmark found that pneumonia (ICD10: J18) was the single diagnosis associated with most deaths within 30 days following any hospital admission, exceeding deaths from acute myocardial infarction, stroke, heart failure, and lung cancer.13 To elucidate the magnitude of infection-related mortality, several limitations of cause of death statistics should be considered. First, the accuracy of cause of death data is limited both in Denmark and elsewhere due to low autopsy rates and even deliberate, erroneous coding.14-17 Second, to conceptualize 1 single underlying cause of death is complex in patients with several coexisting diseases.18 Third, information on contributory causes of death are
0196-6553/$36.00 - Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2013.12.011
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
often lacking in the statistics. Fourth, infection diagnosis codes are scattered throughout organ disease-specific ICD-10 chapters. Knowledge of the prevalence of infection-related hospital death is crucial for planning preventive and therapeutic strategies. We hypothesized that infections are much more prevalent among patients dying in internal medicine departments than apparent from current statistics, and may constitute an under-registered contributory cause of death. With this study, we aimed to estimate the prevalence of infection among people dying in internal medicine departments at a large university hospital and to assess the validity and completeness of the Danish CDR and HDR for infection-related deaths.
507
Identification of patients with infection during terminal hospital admission We used the Danish Civil Registration System to identify all Danish citizens who died in Denmark during 2008. Civil registry numbers were used to link this information with hospital discharge diagnosis codes in the HDR, thereby identifying all patients who died during admission to an internal medicine department at Aalborg University Hospital during 2008. Study subjects were patients with any ICD-10 infectious disease diagnosis code present either in the CDR or in the HDR in relation to the admission terminated by death (including but not limited to ICD-10: A00-B99, respiratory tract infections from the J block, urinary tract infections from the N block, and postprocedural infections from the T block).
METHODS The Aalborg University Hospital microbiology register Setting This study took place in the North Denmark Region (w580,000 inhabitants) and drew on prospectively recorded data within the period January 1, 2008-December 31, 2008. We choose 2008 as the study year because both publication of CDR data and scanned electronic copies of the medical records of deceased patients are currently delayed by several years. We identified all adult patients (aged 15 years) who died during hospitalization at 1 of Aalborg University Hospital’s internal medicine departments. Eleven departments were included and the medical specialties represented at these departments were cardiology, rheumatology, endocrinology, hematology, lung medicine, renal medicine, neurology, oncology, gastroenterology, infectious medicine, and geriatric medicine. Aalborg University Hospital comprises approximately 700 beds and serves as both a district hospital with an admission area of 250,000 inhabitants and as a referral hospital for the North Denmark Region. All residents of Denmark are assigned a unique personal identifier, the civil registry number, at birth or immigration. This enabled us to link data from multiple health care databases and medical records for this study.19
HDR The HDR comprises data on all patients discharged from somatic departments since 1977, including both administrative and clinical data. The administrative data includes information on patient identification, time of admission and discharge, and identification of hospital ward. Each hospital discharge is recorded in the registry with 1 primary (ie, first-listed) diagnosis and up to 20 secondary discharge diagnoses coded by physicians in the department since 1994 according to the ICD-10.20
CDR The CDR is maintained by the Danish National Board of Health and contains individual-level data on all deaths among Danish residents dying in Denmark. Since 1994 causes of death have been classified in accordance with the ICD-10. The registry contains information from electronic death certificates issued by the medical doctor who verified the death. The death certificate states mandatory information on the underlying cause of death defined as the condition, disease, or event initiating and being a precondition for the inevitable sequence of events leading to death. In addition, the death certificate includes optional information on contributing causes of death (up to 2 intermediate causes of death and the immediate cause of death).21
The Department of Clinical Microbiology, Aalborg University Hospital, provides bacteriology services for the entire county. All microbiology information is recorded in the computerized laboratory information system (ADBakt, Autonik, Ramstra, Sköldinge, Sweden) from which we retrieved information on microbiology findings. Medical record validation of infection All records were reviewed by 1 reviewer (ABB) and discussed with a senior author (MDP) in any case of doubt. The records were examined with respect to the criteria of infectious disease stated below. We considered infection to be verified if 1 of the following criteria were satisfied: Any 2 of the following 3: An infectious disease diagnosis made by a treating physician and registered in the medical record, a prescription of antimicrobial therapy, or a positive microbiology result from a relevant site (consistent with the site considered as focus by the treating physician). A positive microbiology culture from a normally sterile body site (eg, blood, cerebrospinal fluid, pleural fluid, ascetic fluid, or synovial fluid). Active chronic infectious disease (eg, HIV/AIDS, hepatitis B or C, or tuberculosis). Infectious disease verified at autopsy. Because our primary goal was to evaluate the prevalence of infections as a possible underlying or contributory cause of death, we distinguished between major and minor infections before death. We a priori defined major infections as bacterial infections of the lower respiratory tract, upper urinary tract, skin, bone, joint, internal abdomen, central nervous system, and cardiovascular system. This category also included chronic infectious diseases, including tuberculosis, HIV, and viral hepatitis. Episodes of bacteremia and fungemia were also considered major infections (irrespective of focus) if contamination was ruled out by senior physicians at the regional Department of Clinical Microbiology. Oral candidiasis and bacterial cystitis are important examples of infections that we considered minor. Because some patients experienced prolonged hospital admission before death, we furthermore evaluated if patients had an ongoing infection during their final 7 days of life. Statistical analyses We examined the validity of an infection diagnosis separately in the CDR and the HDR. The validity was expressed as the positive
508
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
Table 1 Positive predictive value (PPV) of infection diagnosis as a contributory cause of death for the Cause of Death Registry (CDR) and the Hospital Discharge Registry (HDR) and completeness of infectious disease registration in the CDR and HDR No. of cases with verified infection/No. of cases in register CDR Underlying cause of death Underlying or contributing cause of death HDR
31/31
PPV (95% CI) 100.00 (92.3-1.0)
Completeness (95% CI) ...
145/151
96.0 (92.0-98.3) 79.2 (72.9-84.6)
141/148
95.3 (90.9-97.9) 77.0 (70.6-82.7)
CI, confidence interval. NOTE. Values for PPV and completeness are given as %.
Fig 1. Nonproportional Venn diagram illustrating the registration of 193 patients who died during hospitalization in internal medicine departments and who had an infection diagnosis code recorded either in the Cause of Death Registry or the Hospital Discharge Registry (primary or secondary diagnosis). Among the 151 patients registered in the Cause of Death Registry, 31 patients (16.1%) had infection recorded as the underlying cause of death (not shown).
predictive value (PPV) of an infection diagnosis code, calculated as the proportion of patients with an infectious disease diagnosis in the CDR and/or the HDR who had an infection according to their medical record. Moreover, we estimated the completeness of the CDR and the HDR as the proportion of the total number of verified infections in our study registered in the respective register (Table 1). In addition, we specifically estimated the PPV of a lower respiratory tract infection code (the most common infectious focus) in the CDR and the HDR for verified lower respiratory tract infection. We used Jeffrey’s method to estimate 95% confidence intervals (CIs) for PPV and completeness.22 Finally, we used 2-sample captureerecapture analysis to estimate the total number of infectious disease-related deaths and the number of patients missed by both information systems.23 STATA statistical software, version 11.2 (StataCorp, College Station, TX) was used for all analyses. The study was approved by the Danish Data Protection Agency (record No. 2006-53-1396). RESULTS A total of 458 adult patients died during their stay in an internal medicine department at Aalborg University Hospital in 2008. Of those, 193 patients (42.1%) had an infection diagnosis code
recorded either in the HDR or the CDR. The registration of these patients is shown in Figure 1. The median age of patients who died with infection was 77.1 years (interquartile range, 69.3-84.4 years) and the median length of stay was 6 days (interquartile range, 2-15 days). Table 2 shows the extent to which the predetermined criteria were fulfilled among the 193 patients with an infection diagnosis. According to this, 94.8% of patients (183 of 193) had verified infection and 74.1% of patients (143 of 193) were classified as having a major infection. Furthermore, 88.1% of patients (170 of 193) had signs of ongoing infection during the final 7 days of life. The major types of infection recorded in the medical records by the treating physician during the terminal hospitalization are shown in Table 2. The infection most frequently related to death was pneumonia (55.4%; 107 of 193) followed by infections with unknown (23.3%; 45 of 193) and abdominal focus (11.4%; 22 of 193). Upper and lower urinary tract infections were recorded in 5.2% of patients (10 of 193) and 7.8% of patients (15 of 193), respectively. The PPV of an infectious disease diagnosis in the CDR was 96.0% (145 of 151; 95% CI, 92.0%-98.3%) when considering the underlying and contributory causes of death listed and 100.00% (31 of 31; 95% CI, 92.3%-100.0%) when limited to the underlying causes of death listed (Table 1). The PPV of an infectious disease diagnosis in the HDR was 95.3% (141 of 148; 95% CI, 90.9%-97.9%). The PPV of a lower respiratory tract infection code in the HDR versus verified lower respiratory tract infection was 78.3% (95% CI, 69.8%-85.3%) (primary diagnosis, 76.9% [95% CI, 65.6%-85.9%] and secondary diagnosis, 81.8% [95% CI, 68.6%-91.0%]) and for the CDR it was 74.3% (95% CI, 65.3%-81.9%) (underlying cause of death, 88.9% [95% CI, 68.9%97.6%] and contributory cause of death, 73.3% [95% CI, 63.6%81.6%]). The completeness of the CDR for verified infection was 79.2% [95% CI, 72.9%-84.6%] when considering both underlying and contributory causes of death. A similar degree of completeness was found for the HDR (77.0% [95% CI, 70.6%-82.7%]). By capturee recapture analysis we estimated that the overall number of infection-related deaths was 210. When we restricted the captureerecapture analysis to patients with verified infection we estimated 198 true infection-related deaths (43.2% of all 458 deaths in internal medicine departments) corresponding to 47 and 43 cases missed by the HDR and CDR, respectively. DISCUSSION Our study revealed a very high prevalence of infections among patients who died during hospitalization in university hospital internal medicine departments. We found a high PPV for infectious disease diagnoses in both the HDR and the CDR; however, the level of completeness for infection in each registry was only modest. Strengths of our study include a high degree of population coverage of both health care registries evaluated. The Danish National Board of Health asserts that the CDR covers 99.4%-99.7% of all deaths of Danish citizens dying in Denmark, whereas the HDR internationally is considered to be 1 of the most comprehensive population-based databases of its kind.20,21 A number of limitations must be considered when interpreting our results. First, this was a single-center study. Although our study population was restricted to patients dying at 1 large university hospital, we believe the results can be generalized to patients admitted to internal medicine departments at other Danish hospitals. The secondary health care sector in Denmark consists of approximately 40 public hospitals, and the Danish National Health Service guarantees unfettered access free of charge at the point of delivery at all these hospitals. At Aalborg University Hospital almost all medical specialties are present, including departments for
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
advanced treatment of cancer, immunosuppressive disorders, and infectious diseases, which may have increased the overall prevalence of infections in our study. A recent Danish report states that 48% of all Danish residents died at the hospital during the period 2004-2006.24 The report did not asses the proportion of these deaths that took place in internal medicine departments; however, we expect it to be the majority. Our study revealed a 40% prevalence of infections among patients who died at internal medicine departments and in >75% of these cases the infection was considered major. If these results can be generalized to patients admitted to other Danish hospitals, a large proportion of the Danish population is consequently likely to die from infection-related death, indicating that the annual reports on causes of death greatly underestimate infections as a contributory cause of death. Another limitation is that our study period was limited to 1 year. However, given that the disease-scape is unlikely to have changed substantially since 2008 in Denmark and that the registration and reporting of diseases to the CDR and HDR have been unchanged, we believe the findings to be representative of later years. Furthermore, we only evaluated 1 dimension of validity, namely the PPV of a recorded infection in association with death. Our study did not include evaluation of sensitivity and specificity of the registries that would have required detailed examination of all episodes of death. Moreover, the exact quality of information recorded and abstracted from the medical records may be difficult to assess. Diagnoses of clinical infection in the medical records made by the treating physician without positive microbiology results may have been made on an uncertain basis and do not necessarily equal true infection. This would lead to an overestimation of the total number of infection-related deaths. Finally, the 2-sample method captureerecapture analysis we used was based on the assumption that the registries are independent. This may not be fulfilled in our setting; for instance, if the same doctor reports to both registries. However, such dependence between the 2 registers would most likely result in a capturee recapture estimate biased toward an underestimate.25 Studies on overall infection-related death are sparse. However, national statistics offices worldwide deliver information on underlying causes of deaths with little or no information on contributory causes of death. The Danish Ministry of Health reports that infections cause only 1.6% of all deaths, although a rather arbitrary definition of infectious diseases is used that is limited to ICD-10: A00-B99.12 In the United States, influenza and pneumonia (ICD10: J09-J18) is the ninth and septicemia the 11th leading cause of death.26 Furthermore, the World Health Organization reports that in high-income countries lower respiratory infection is the fifth leading cause of death accounting for 3.8% of all deaths or 350,000 deaths a year.11 Still, due to the focus on potential underlying causes of death, these data do not fully illustrate the burden of infectionrelated death, as illustrated by Christensen et al,13 who found pneumonia to be the number-1 diagnosis associated with death within 30 days of hospitalization. Two problems with the use of only 1 underlying cause of death should be considered. First, it may be difficult to identify 1 disease as the underlying cause of death. It does not require a vivid imagination to conjure up a situation where this is the case. An octogenarian with diabetes mellitus, heart failure, and chronic obstructive pulmonary disease dies after a few days of pneumococcal pneumonia; not an unusual case in high-income countries. In the European region >70% of all mortality occurs in people aged 65 years or older,27 and in Denmark the proportion is even higher, exceeding 80%.12 Overall, the prevalence of multimorbidity has been estimated at 25% in Scotland,28 and for people older than age 65 years the prevalence is 55%98%.28,29 Thus, comorbidity is common in people dying in highincome countries, and the challenge of identifying 1 underlying cause of death among many is also common.
509
Table 2 Infections among all 193 adults who died while hospitalized in internal medicine departments at Aalborg University Hospital during 2008 and later had an infection diagnosis code registered in the Cause of Death Registry (CDR) and/or the Hospital Discharge Registry (HDR) Patients Infection Infection the underlying cause of death in CDR Infection in CDR (underlying and/or contributing) Infection in HDR Infection in both CDR and HDR Infection in CDR but not in HDR Infection in HDR but not in CDR Source of infection according to treating physician All Lower respiratory tract Unknown source Abdomen Lower urinary tract Upper urinary tract Other source Antimicrobial treatment initiated Any positive culture from the microbiology laboratory Positive culture (sterile and/or unsterile site) Positive culture from sterile site Positive blood culture Other positive microbiology culture Infection according to investigator definitions Major infection >1 infection according to medical files Infection active during the final 7 d of life Length of stay 7 d
n
%
31 151 148 106 45 42
16.1 78.2 76.7 54.9 23.3 21.8
181 107 45 22 15 10 13 184 107 102 36 33 12 183 143 32 170* 111
93.8 55.4 23.3 11.4 7.8 5.2 6.7 94.3 55.4 52.9 18.7 17.1 6.2 94.8 74.1 16.6 88.1 57.5
*During the final 7 days of life 15 patients did not have active signs of ongoing infection, whereas active infection was deemed uncertain for 9 patients.
Second, the use of only 1 underlying cause of death in older adults may tend to overestimate certain chronic diseases. Indeed, a recent study by Tinetti et al18 found that cancer and cardiovascular diseases may be overrepresented as causes of death. Furthermore, in a recent survey of physicians from the United States, Wexelman et al17 found that half of residents had knowingly reported an inaccurate cause of death, which was predominantly a cardiovascular disease. The study by Wexelman et al17 highlights another important limitation of cause of death statistics, namely the modest reliability of cause of death certificates.14-16
CONCLUSIONS We found a very high prevalence of infectious diseases among patients who died during admission to internal medicine departments. Our data suggest that the current practice of determining only 1 underlying cause of death seems to misrepresent the burden of infection-related death. We believe these prevalence data are important and emphasize the need for optimized infection prevention and treatment strategies. The high PPV of infection diagnoses estimated in our study indicates a low false-positive number of infectious disease cases in both the CDR and the HDR, suggesting that these registries are valid sources for identifying cases of infectious diseases for epidemiologic research. However, separately the registries were incomplete and need mutual complementation when estimating the total number of infectious disease-caused deaths.
Acknowledgment The authors thank Elisabeth Kristoffersen for helping to retrieve and systemize the medical records.
510
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
References 1. Baker MG, Barnard LT, Kvalsvig A, Verrall A, Zhang J, Keall M, et al. Increasing incidence of serious infectious diseases and inequalities in New Zealand: a national epidemiological study. Lancet 2012;379:1112-9. 2. Christensen KL, Holman RC, Steiner CA, Sejvar JJ, Stoll BJ, Schonberger LB. Infectious disease hospitalizations in the United States. Clin Infect Dis 2009;49: 1025-35. 3. Elixhauser A, Owens P. Reasons for being admitted to the hospital through the emergency department, 2003. Available from: http://www.hcup-us.ahrq.gov/ reports/statbriefs/sb2.pdf. Updated 2006. Accessed June 10, 2013. 4. Etyang AO, Scott JA. Medical causes of admissions to hospital among adults in Africa: a systematic review. Glob Health Action 2013;6:1-14. 5. Thomsen RW, Riis A, Norgaard M, Jacobsen J, Christensen S, McDonald CJ, et al. Rising incidence and persistently high mortality of hospitalized pneumonia: a 10-year population-based study in Denmark. J Intern Med 2006;259:410-7. 6. Sogaard M, Norgaard M, Dethlefsen C, Schonheyder HC. Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study. Clin Infect Dis 2011;52:61-9. 7. Melamed A, Sorvillo FJ. The burden of sepsis-associated mortality in the United States from 1999 to 2005: an analysis of multiple-cause-of-death data. Crit Care 2009;13:R28. 8. Dombrovskiy VY, Martin AA, Sunderram J, Paz HL. Rapid increase in hospitalization and mortality rates for severe sepsis in the United States: a trend analysis from 1993 to 2003. Crit Care Med 2007;35:1244-50. 9. Burke JP. Infection control - a problem for patient safety. N Engl J Med 2003; 348:651-6. 10. Klausen HH, Petersen J, Lindhardt T, Bandholm T, Hendricksen C, Kehlet H, et al. Outcomes in elderly Danish citizens admitted with community-acquired pneumonia. Regional differences in a public healthcare system. Respir Med 2012;106:1778-87. 11. The top 10 causes of death. Available from: http://who.int/mediacentre/ factsheets/fs310/en/index.html. Updated 2011. Accessed May 19, 2013. 12. Tal og analyse: dødsårsagsregistret 2011. Available from: http://www.ssi.dk/ Sundhedsdataogit/Registre/w/media/Indhold/DK%20-%20dansk/Sundhedsdata% 20og%20it/NSF/Registre/Dodsaarsagsregisteret/D%C3%B8ds%C3%A5rsagsregisteret %202011.ashx. Published 2011. Accessed June 12, 2013. 13. Christensen S, Jacobsen J, Bartels P, Norgaard M. Hospital standardised mortality ratios based on data from administrative registries. A pilot project. Ugeskr Laeger 2007;169:2767-72.
14. Jensen HH, Godtfredsen NS, Lange P, Vestbo J. Potential misclassification of causes of death from COPD. Eur Respir J 2006;28:781-5. 15. Sidenius KE, Munch EP, Madsen FF, Lange P, Viskum K, Soes-Petersen U. Reliability of recorded deaths from asthma in Denmark during a 1-year period 1994-1995. Ugeskr Laeger 2001;163:6131-3. 16. Gjersoe P, Andersen SE, Molbak AG, Wulff HR, Thomsen OO. Reliability of death certificates. The reproducibility of the recorded causes of death in patients admitted to departments of internal medicine. Ugeskr Laeger 1998;160:5030-4. 17. Wexelman BA, Eden E, Rose KM. Survey of New York City resident physicians on cause-of-death reporting, 2010. Prev Chronic Dis 2013;10:E76. 18. Tinetti ME, McAvay GJ, Murphy TE, Gross CP, Lin H, Allore HG. Contribution of individual diseases to death in older adults with multiple diseases. J Am Geriatr Soc 2012;60:1448-56. 19. Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011; 39(7 Suppl):22-5. 20. Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health 2011;39(7 Suppl):30-3. 21. Helweg-Larsen K. The Danish Register of Causes of Death. Scand J Public Health 2011;39(7 Suppl):26-9. 22. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Stat Sci 2001;16:101-33. 23. Chao A, Tsay PK, Lin SH, Shau WY, Chao DY. The applications of capturerecapture models to epidemiological data. Stat Med 2001;20:3123-57. 24. Dødssted i Danmark i perioden 2004 til 2006. Available from: http://pavi.dk/ Libraries/projekter/Doedssted_i_Danmark_i_perioden_2004-2006.sflb.ashx. Published 2013. Accessed June 12, 2013. 25. Hook EB, Regal RR. Accuracy of alternative approaches to capture-recapture estimates of disease frequency: internal validity analysis of data from five sources. Am J Epidemiol 2000;152:771-9. 26. Hoyert D, Xu J. Deaths: preliminary data for 2011. Natl Vital Stat Rep 2012;61: 1-52. 27. The European health report 2012: charting the way to well-being. Available from: http://www.euro.who.int/en/what-we-do/data-and-evidence/european-healthreport-2012/the-european-health-report-2012-charting-the-way-to-well-being. Published 2013. Accessed June 12, 2013. 28. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet 2012;380:37-43. 29. Marengoni A, Angleman S, Melis R, Mangialasche F, Karp A, Garmen A, et al. Aging with multimorbidity: a systematic review of the literature. Ageing Res Rev 2011;10:430-9.
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
Relationship between death and infections among patients hospitalized in internal medicine departments: A prevalence and validation study Anne B. Boisen MB a, Michael Dalager-Pedersen MD a, b, Mette Søgaard DVM, PhD a, Rikke Mortensen MSc a, Reimar W. Thomsen MD, PhD a, * a b
Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Aarhus, Denmark Department of Infectious Diseases, Aalborg University Hospital, Aalborg, Denmark
Key Words: Bacterial infections and mycoses Infectious disease medicine Cause of death/vital statistics Epidemiology
Background: We assessed the burden of infection-related death at internal medicine departments in Denmark and the validity of 2 population-based registries for diagnoses of infection. Methods: We reviewed medical records of all inpatients who died at a large university hospital during 2008 with an infection diagnosis in the Cause of Death Registry (CDR) or Hospital Discharge Registry (HDR). We computed the positive predictive value of infection diagnoses and completeness of each registry with 95% confidence intervals (CIs) and the prevalence of infection-related deaths by capturerecapture analysis. Results: Among 458 deaths, 193 patients (42.1%) had an infection diagnosis of which 40.0% (183 out of 458) were verified. The positive predictive value of an infection diagnosis was 96.0% (95% CI, 92.0%98.3%) in the CDR and 95.3% (CI, 90.9%-97.9%) in the HDR. Completeness of the CDR was 79.2% (CI, 72.9%84.6%) and completeness of the HDR was 77.0% (CI, 70.6%-82.7%). By captureerecapture analysis we estimated that 43.2% of all deaths (198 out of 458) were related to infection. Conclusions: The large proportion of deaths in internal medicine departments that are related to infection emphasizes the need for optimized infection prevention and treatment strategies. CDR and HDR are valid sources for identifying infection and may supplement each other for increased completeness of infection-related death in epidemiologic research. Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Infectious diseases are among the most frequent causes of acute hospital admission both in low- and high-income countries.1-4 During past decades, hospitalization rates for severe infections such as pneumonia or bacteremia/sepsis have increased in the Western world.5-8 Moreover, approximately 10% of patients acutely admitted to a hospital acquire 1 infections during their hospital stay.9 The associated in-hospital mortality approaches 4% following hospitalization with any infection1 and 10%-15% for patients with pneumonia or bacteremia.5,6,10 As a result, a considerable proportion of all deaths are expected to be related to infections. In highincome countries, cancer and heart diseases are the most
* Address correspondence to Reimar W. Thomsen, MD, PhD, Department of Clinical Epidemiology, Institute of Clinical Medicine, Aarhus University Hospital, Olof Palmes Allé 43-45, DK-8200 Aarhus N, Denmark. E-mail address: [email protected] (R.W. Thomsen). Funding for this investigation was provided by Aalborg University Hospital. Conflicts of interest: None to report.
common underlying causes of death.11 According to data from the Danish Cause of Death Registry (CDR), cancer accounted for 30.4% and heart disease for 16.9% of all deaths in Denmark during 2011. In contrast, an infectious disease as coded in ICD-10 was the underlying cause in only 1.6% of deaths.12 Notwithstanding, a recent population-based study using Hospital Discharge Registry (HDR) data from northern Denmark found that pneumonia (ICD10: J18) was the single diagnosis associated with most deaths within 30 days following any hospital admission, exceeding deaths from acute myocardial infarction, stroke, heart failure, and lung cancer.13 To elucidate the magnitude of infection-related mortality, several limitations of cause of death statistics should be considered. First, the accuracy of cause of death data is limited both in Denmark and elsewhere due to low autopsy rates and even deliberate, erroneous coding.14-17 Second, to conceptualize 1 single underlying cause of death is complex in patients with several coexisting diseases.18 Third, information on contributory causes of death are
0196-6553/$36.00 - Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2013.12.011
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
often lacking in the statistics. Fourth, infection diagnosis codes are scattered throughout organ disease-specific ICD-10 chapters. Knowledge of the prevalence of infection-related hospital death is crucial for planning preventive and therapeutic strategies. We hypothesized that infections are much more prevalent among patients dying in internal medicine departments than apparent from current statistics, and may constitute an under-registered contributory cause of death. With this study, we aimed to estimate the prevalence of infection among people dying in internal medicine departments at a large university hospital and to assess the validity and completeness of the Danish CDR and HDR for infection-related deaths.
507
Identification of patients with infection during terminal hospital admission We used the Danish Civil Registration System to identify all Danish citizens who died in Denmark during 2008. Civil registry numbers were used to link this information with hospital discharge diagnosis codes in the HDR, thereby identifying all patients who died during admission to an internal medicine department at Aalborg University Hospital during 2008. Study subjects were patients with any ICD-10 infectious disease diagnosis code present either in the CDR or in the HDR in relation to the admission terminated by death (including but not limited to ICD-10: A00-B99, respiratory tract infections from the J block, urinary tract infections from the N block, and postprocedural infections from the T block).
METHODS The Aalborg University Hospital microbiology register Setting This study took place in the North Denmark Region (w580,000 inhabitants) and drew on prospectively recorded data within the period January 1, 2008-December 31, 2008. We choose 2008 as the study year because both publication of CDR data and scanned electronic copies of the medical records of deceased patients are currently delayed by several years. We identified all adult patients (aged 15 years) who died during hospitalization at 1 of Aalborg University Hospital’s internal medicine departments. Eleven departments were included and the medical specialties represented at these departments were cardiology, rheumatology, endocrinology, hematology, lung medicine, renal medicine, neurology, oncology, gastroenterology, infectious medicine, and geriatric medicine. Aalborg University Hospital comprises approximately 700 beds and serves as both a district hospital with an admission area of 250,000 inhabitants and as a referral hospital for the North Denmark Region. All residents of Denmark are assigned a unique personal identifier, the civil registry number, at birth or immigration. This enabled us to link data from multiple health care databases and medical records for this study.19
HDR The HDR comprises data on all patients discharged from somatic departments since 1977, including both administrative and clinical data. The administrative data includes information on patient identification, time of admission and discharge, and identification of hospital ward. Each hospital discharge is recorded in the registry with 1 primary (ie, first-listed) diagnosis and up to 20 secondary discharge diagnoses coded by physicians in the department since 1994 according to the ICD-10.20
CDR The CDR is maintained by the Danish National Board of Health and contains individual-level data on all deaths among Danish residents dying in Denmark. Since 1994 causes of death have been classified in accordance with the ICD-10. The registry contains information from electronic death certificates issued by the medical doctor who verified the death. The death certificate states mandatory information on the underlying cause of death defined as the condition, disease, or event initiating and being a precondition for the inevitable sequence of events leading to death. In addition, the death certificate includes optional information on contributing causes of death (up to 2 intermediate causes of death and the immediate cause of death).21
The Department of Clinical Microbiology, Aalborg University Hospital, provides bacteriology services for the entire county. All microbiology information is recorded in the computerized laboratory information system (ADBakt, Autonik, Ramstra, Sköldinge, Sweden) from which we retrieved information on microbiology findings. Medical record validation of infection All records were reviewed by 1 reviewer (ABB) and discussed with a senior author (MDP) in any case of doubt. The records were examined with respect to the criteria of infectious disease stated below. We considered infection to be verified if 1 of the following criteria were satisfied: Any 2 of the following 3: An infectious disease diagnosis made by a treating physician and registered in the medical record, a prescription of antimicrobial therapy, or a positive microbiology result from a relevant site (consistent with the site considered as focus by the treating physician). A positive microbiology culture from a normally sterile body site (eg, blood, cerebrospinal fluid, pleural fluid, ascetic fluid, or synovial fluid). Active chronic infectious disease (eg, HIV/AIDS, hepatitis B or C, or tuberculosis). Infectious disease verified at autopsy. Because our primary goal was to evaluate the prevalence of infections as a possible underlying or contributory cause of death, we distinguished between major and minor infections before death. We a priori defined major infections as bacterial infections of the lower respiratory tract, upper urinary tract, skin, bone, joint, internal abdomen, central nervous system, and cardiovascular system. This category also included chronic infectious diseases, including tuberculosis, HIV, and viral hepatitis. Episodes of bacteremia and fungemia were also considered major infections (irrespective of focus) if contamination was ruled out by senior physicians at the regional Department of Clinical Microbiology. Oral candidiasis and bacterial cystitis are important examples of infections that we considered minor. Because some patients experienced prolonged hospital admission before death, we furthermore evaluated if patients had an ongoing infection during their final 7 days of life. Statistical analyses We examined the validity of an infection diagnosis separately in the CDR and the HDR. The validity was expressed as the positive
508
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
Table 1 Positive predictive value (PPV) of infection diagnosis as a contributory cause of death for the Cause of Death Registry (CDR) and the Hospital Discharge Registry (HDR) and completeness of infectious disease registration in the CDR and HDR No. of cases with verified infection/No. of cases in register CDR Underlying cause of death Underlying or contributing cause of death HDR
31/31
PPV (95% CI) 100.00 (92.3-1.0)
Completeness (95% CI) ...
145/151
96.0 (92.0-98.3) 79.2 (72.9-84.6)
141/148
95.3 (90.9-97.9) 77.0 (70.6-82.7)
CI, confidence interval. NOTE. Values for PPV and completeness are given as %.
Fig 1. Nonproportional Venn diagram illustrating the registration of 193 patients who died during hospitalization in internal medicine departments and who had an infection diagnosis code recorded either in the Cause of Death Registry or the Hospital Discharge Registry (primary or secondary diagnosis). Among the 151 patients registered in the Cause of Death Registry, 31 patients (16.1%) had infection recorded as the underlying cause of death (not shown).
predictive value (PPV) of an infection diagnosis code, calculated as the proportion of patients with an infectious disease diagnosis in the CDR and/or the HDR who had an infection according to their medical record. Moreover, we estimated the completeness of the CDR and the HDR as the proportion of the total number of verified infections in our study registered in the respective register (Table 1). In addition, we specifically estimated the PPV of a lower respiratory tract infection code (the most common infectious focus) in the CDR and the HDR for verified lower respiratory tract infection. We used Jeffrey’s method to estimate 95% confidence intervals (CIs) for PPV and completeness.22 Finally, we used 2-sample captureerecapture analysis to estimate the total number of infectious disease-related deaths and the number of patients missed by both information systems.23 STATA statistical software, version 11.2 (StataCorp, College Station, TX) was used for all analyses. The study was approved by the Danish Data Protection Agency (record No. 2006-53-1396). RESULTS A total of 458 adult patients died during their stay in an internal medicine department at Aalborg University Hospital in 2008. Of those, 193 patients (42.1%) had an infection diagnosis code
recorded either in the HDR or the CDR. The registration of these patients is shown in Figure 1. The median age of patients who died with infection was 77.1 years (interquartile range, 69.3-84.4 years) and the median length of stay was 6 days (interquartile range, 2-15 days). Table 2 shows the extent to which the predetermined criteria were fulfilled among the 193 patients with an infection diagnosis. According to this, 94.8% of patients (183 of 193) had verified infection and 74.1% of patients (143 of 193) were classified as having a major infection. Furthermore, 88.1% of patients (170 of 193) had signs of ongoing infection during the final 7 days of life. The major types of infection recorded in the medical records by the treating physician during the terminal hospitalization are shown in Table 2. The infection most frequently related to death was pneumonia (55.4%; 107 of 193) followed by infections with unknown (23.3%; 45 of 193) and abdominal focus (11.4%; 22 of 193). Upper and lower urinary tract infections were recorded in 5.2% of patients (10 of 193) and 7.8% of patients (15 of 193), respectively. The PPV of an infectious disease diagnosis in the CDR was 96.0% (145 of 151; 95% CI, 92.0%-98.3%) when considering the underlying and contributory causes of death listed and 100.00% (31 of 31; 95% CI, 92.3%-100.0%) when limited to the underlying causes of death listed (Table 1). The PPV of an infectious disease diagnosis in the HDR was 95.3% (141 of 148; 95% CI, 90.9%-97.9%). The PPV of a lower respiratory tract infection code in the HDR versus verified lower respiratory tract infection was 78.3% (95% CI, 69.8%-85.3%) (primary diagnosis, 76.9% [95% CI, 65.6%-85.9%] and secondary diagnosis, 81.8% [95% CI, 68.6%-91.0%]) and for the CDR it was 74.3% (95% CI, 65.3%-81.9%) (underlying cause of death, 88.9% [95% CI, 68.9%97.6%] and contributory cause of death, 73.3% [95% CI, 63.6%81.6%]). The completeness of the CDR for verified infection was 79.2% [95% CI, 72.9%-84.6%] when considering both underlying and contributory causes of death. A similar degree of completeness was found for the HDR (77.0% [95% CI, 70.6%-82.7%]). By capturee recapture analysis we estimated that the overall number of infection-related deaths was 210. When we restricted the captureerecapture analysis to patients with verified infection we estimated 198 true infection-related deaths (43.2% of all 458 deaths in internal medicine departments) corresponding to 47 and 43 cases missed by the HDR and CDR, respectively. DISCUSSION Our study revealed a very high prevalence of infections among patients who died during hospitalization in university hospital internal medicine departments. We found a high PPV for infectious disease diagnoses in both the HDR and the CDR; however, the level of completeness for infection in each registry was only modest. Strengths of our study include a high degree of population coverage of both health care registries evaluated. The Danish National Board of Health asserts that the CDR covers 99.4%-99.7% of all deaths of Danish citizens dying in Denmark, whereas the HDR internationally is considered to be 1 of the most comprehensive population-based databases of its kind.20,21 A number of limitations must be considered when interpreting our results. First, this was a single-center study. Although our study population was restricted to patients dying at 1 large university hospital, we believe the results can be generalized to patients admitted to internal medicine departments at other Danish hospitals. The secondary health care sector in Denmark consists of approximately 40 public hospitals, and the Danish National Health Service guarantees unfettered access free of charge at the point of delivery at all these hospitals. At Aalborg University Hospital almost all medical specialties are present, including departments for
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
advanced treatment of cancer, immunosuppressive disorders, and infectious diseases, which may have increased the overall prevalence of infections in our study. A recent Danish report states that 48% of all Danish residents died at the hospital during the period 2004-2006.24 The report did not asses the proportion of these deaths that took place in internal medicine departments; however, we expect it to be the majority. Our study revealed a 40% prevalence of infections among patients who died at internal medicine departments and in >75% of these cases the infection was considered major. If these results can be generalized to patients admitted to other Danish hospitals, a large proportion of the Danish population is consequently likely to die from infection-related death, indicating that the annual reports on causes of death greatly underestimate infections as a contributory cause of death. Another limitation is that our study period was limited to 1 year. However, given that the disease-scape is unlikely to have changed substantially since 2008 in Denmark and that the registration and reporting of diseases to the CDR and HDR have been unchanged, we believe the findings to be representative of later years. Furthermore, we only evaluated 1 dimension of validity, namely the PPV of a recorded infection in association with death. Our study did not include evaluation of sensitivity and specificity of the registries that would have required detailed examination of all episodes of death. Moreover, the exact quality of information recorded and abstracted from the medical records may be difficult to assess. Diagnoses of clinical infection in the medical records made by the treating physician without positive microbiology results may have been made on an uncertain basis and do not necessarily equal true infection. This would lead to an overestimation of the total number of infection-related deaths. Finally, the 2-sample method captureerecapture analysis we used was based on the assumption that the registries are independent. This may not be fulfilled in our setting; for instance, if the same doctor reports to both registries. However, such dependence between the 2 registers would most likely result in a capturee recapture estimate biased toward an underestimate.25 Studies on overall infection-related death are sparse. However, national statistics offices worldwide deliver information on underlying causes of deaths with little or no information on contributory causes of death. The Danish Ministry of Health reports that infections cause only 1.6% of all deaths, although a rather arbitrary definition of infectious diseases is used that is limited to ICD-10: A00-B99.12 In the United States, influenza and pneumonia (ICD10: J09-J18) is the ninth and septicemia the 11th leading cause of death.26 Furthermore, the World Health Organization reports that in high-income countries lower respiratory infection is the fifth leading cause of death accounting for 3.8% of all deaths or 350,000 deaths a year.11 Still, due to the focus on potential underlying causes of death, these data do not fully illustrate the burden of infectionrelated death, as illustrated by Christensen et al,13 who found pneumonia to be the number-1 diagnosis associated with death within 30 days of hospitalization. Two problems with the use of only 1 underlying cause of death should be considered. First, it may be difficult to identify 1 disease as the underlying cause of death. It does not require a vivid imagination to conjure up a situation where this is the case. An octogenarian with diabetes mellitus, heart failure, and chronic obstructive pulmonary disease dies after a few days of pneumococcal pneumonia; not an unusual case in high-income countries. In the European region >70% of all mortality occurs in people aged 65 years or older,27 and in Denmark the proportion is even higher, exceeding 80%.12 Overall, the prevalence of multimorbidity has been estimated at 25% in Scotland,28 and for people older than age 65 years the prevalence is 55%98%.28,29 Thus, comorbidity is common in people dying in highincome countries, and the challenge of identifying 1 underlying cause of death among many is also common.
509
Table 2 Infections among all 193 adults who died while hospitalized in internal medicine departments at Aalborg University Hospital during 2008 and later had an infection diagnosis code registered in the Cause of Death Registry (CDR) and/or the Hospital Discharge Registry (HDR) Patients Infection Infection the underlying cause of death in CDR Infection in CDR (underlying and/or contributing) Infection in HDR Infection in both CDR and HDR Infection in CDR but not in HDR Infection in HDR but not in CDR Source of infection according to treating physician All Lower respiratory tract Unknown source Abdomen Lower urinary tract Upper urinary tract Other source Antimicrobial treatment initiated Any positive culture from the microbiology laboratory Positive culture (sterile and/or unsterile site) Positive culture from sterile site Positive blood culture Other positive microbiology culture Infection according to investigator definitions Major infection >1 infection according to medical files Infection active during the final 7 d of life Length of stay 7 d
n
%
31 151 148 106 45 42
16.1 78.2 76.7 54.9 23.3 21.8
181 107 45 22 15 10 13 184 107 102 36 33 12 183 143 32 170* 111
93.8 55.4 23.3 11.4 7.8 5.2 6.7 94.3 55.4 52.9 18.7 17.1 6.2 94.8 74.1 16.6 88.1 57.5
*During the final 7 days of life 15 patients did not have active signs of ongoing infection, whereas active infection was deemed uncertain for 9 patients.
Second, the use of only 1 underlying cause of death in older adults may tend to overestimate certain chronic diseases. Indeed, a recent study by Tinetti et al18 found that cancer and cardiovascular diseases may be overrepresented as causes of death. Furthermore, in a recent survey of physicians from the United States, Wexelman et al17 found that half of residents had knowingly reported an inaccurate cause of death, which was predominantly a cardiovascular disease. The study by Wexelman et al17 highlights another important limitation of cause of death statistics, namely the modest reliability of cause of death certificates.14-16
CONCLUSIONS We found a very high prevalence of infectious diseases among patients who died during admission to internal medicine departments. Our data suggest that the current practice of determining only 1 underlying cause of death seems to misrepresent the burden of infection-related death. We believe these prevalence data are important and emphasize the need for optimized infection prevention and treatment strategies. The high PPV of infection diagnoses estimated in our study indicates a low false-positive number of infectious disease cases in both the CDR and the HDR, suggesting that these registries are valid sources for identifying cases of infectious diseases for epidemiologic research. However, separately the registries were incomplete and need mutual complementation when estimating the total number of infectious disease-caused deaths.
Acknowledgment The authors thank Elisabeth Kristoffersen for helping to retrieve and systemize the medical records.
510
A.B. Boisen et al. / American Journal of Infection Control 42 (2014) 506-10
References 1. Baker MG, Barnard LT, Kvalsvig A, Verrall A, Zhang J, Keall M, et al. Increasing incidence of serious infectious diseases and inequalities in New Zealand: a national epidemiological study. Lancet 2012;379:1112-9. 2. Christensen KL, Holman RC, Steiner CA, Sejvar JJ, Stoll BJ, Schonberger LB. Infectious disease hospitalizations in the United States. Clin Infect Dis 2009;49: 1025-35. 3. Elixhauser A, Owens P. Reasons for being admitted to the hospital through the emergency department, 2003. Available from: http://www.hcup-us.ahrq.gov/ reports/statbriefs/sb2.pdf. Updated 2006. Accessed June 10, 2013. 4. Etyang AO, Scott JA. Medical causes of admissions to hospital among adults in Africa: a systematic review. Glob Health Action 2013;6:1-14. 5. Thomsen RW, Riis A, Norgaard M, Jacobsen J, Christensen S, McDonald CJ, et al. Rising incidence and persistently high mortality of hospitalized pneumonia: a 10-year population-based study in Denmark. J Intern Med 2006;259:410-7. 6. Sogaard M, Norgaard M, Dethlefsen C, Schonheyder HC. Temporal changes in the incidence and 30-day mortality associated with bacteremia in hospitalized patients from 1992 through 2006: a population-based cohort study. Clin Infect Dis 2011;52:61-9. 7. Melamed A, Sorvillo FJ. The burden of sepsis-associated mortality in the United States from 1999 to 2005: an analysis of multiple-cause-of-death data. Crit Care 2009;13:R28. 8. Dombrovskiy VY, Martin AA, Sunderram J, Paz HL. Rapid increase in hospitalization and mortality rates for severe sepsis in the United States: a trend analysis from 1993 to 2003. Crit Care Med 2007;35:1244-50. 9. Burke JP. Infection control - a problem for patient safety. N Engl J Med 2003; 348:651-6. 10. Klausen HH, Petersen J, Lindhardt T, Bandholm T, Hendricksen C, Kehlet H, et al. Outcomes in elderly Danish citizens admitted with community-acquired pneumonia. Regional differences in a public healthcare system. Respir Med 2012;106:1778-87. 11. The top 10 causes of death. Available from: http://who.int/mediacentre/ factsheets/fs310/en/index.html. Updated 2011. Accessed May 19, 2013. 12. Tal og analyse: dødsårsagsregistret 2011. Available from: http://www.ssi.dk/ Sundhedsdataogit/Registre/w/media/Indhold/DK%20-%20dansk/Sundhedsdata% 20og%20it/NSF/Registre/Dodsaarsagsregisteret/D%C3%B8ds%C3%A5rsagsregisteret %202011.ashx. Published 2011. Accessed June 12, 2013. 13. Christensen S, Jacobsen J, Bartels P, Norgaard M. Hospital standardised mortality ratios based on data from administrative registries. A pilot project. Ugeskr Laeger 2007;169:2767-72.
14. Jensen HH, Godtfredsen NS, Lange P, Vestbo J. Potential misclassification of causes of death from COPD. Eur Respir J 2006;28:781-5. 15. Sidenius KE, Munch EP, Madsen FF, Lange P, Viskum K, Soes-Petersen U. Reliability of recorded deaths from asthma in Denmark during a 1-year period 1994-1995. Ugeskr Laeger 2001;163:6131-3. 16. Gjersoe P, Andersen SE, Molbak AG, Wulff HR, Thomsen OO. Reliability of death certificates. The reproducibility of the recorded causes of death in patients admitted to departments of internal medicine. Ugeskr Laeger 1998;160:5030-4. 17. Wexelman BA, Eden E, Rose KM. Survey of New York City resident physicians on cause-of-death reporting, 2010. Prev Chronic Dis 2013;10:E76. 18. Tinetti ME, McAvay GJ, Murphy TE, Gross CP, Lin H, Allore HG. Contribution of individual diseases to death in older adults with multiple diseases. J Am Geriatr Soc 2012;60:1448-56. 19. Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011; 39(7 Suppl):22-5. 20. Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health 2011;39(7 Suppl):30-3. 21. Helweg-Larsen K. The Danish Register of Causes of Death. Scand J Public Health 2011;39(7 Suppl):26-9. 22. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Stat Sci 2001;16:101-33. 23. Chao A, Tsay PK, Lin SH, Shau WY, Chao DY. The applications of capturerecapture models to epidemiological data. Stat Med 2001;20:3123-57. 24. Dødssted i Danmark i perioden 2004 til 2006. Available from: http://pavi.dk/ Libraries/projekter/Doedssted_i_Danmark_i_perioden_2004-2006.sflb.ashx. Published 2013. Accessed June 12, 2013. 25. Hook EB, Regal RR. Accuracy of alternative approaches to capture-recapture estimates of disease frequency: internal validity analysis of data from five sources. Am J Epidemiol 2000;152:771-9. 26. Hoyert D, Xu J. Deaths: preliminary data for 2011. Natl Vital Stat Rep 2012;61: 1-52. 27. The European health report 2012: charting the way to well-being. Available from: http://www.euro.who.int/en/what-we-do/data-and-evidence/european-healthreport-2012/the-european-health-report-2012-charting-the-way-to-well-being. Published 2013. Accessed June 12, 2013. 28. Barnett K, Mercer SW, Norbury M, Watt G, Wyke S, Guthrie B. Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet 2012;380:37-43. 29. Marengoni A, Angleman S, Melis R, Mangialasche F, Karp A, Garmen A, et al. Aging with multimorbidity: a systematic review of the literature. Ageing Res Rev 2011;10:430-9.