Risk of infections subsequent to pyogenic liver abscess: a nationwide population-based study

Authors


  • Joseph J. Keller and Ming-Chieh Tsai made equal contributions to this manuscript.

Corresponding author: H.-C. Lin, School of Health Care Administration, Taipei Medical University, 250 Wu-Hsing Street, Taipei 110, Taiwan
E-mail: henry11111@tmu.edu.tw

Abstract

This nationwide study aimed to provide risk estimates for a panel of infections subsequent to pyogenic liver abscesses (PLA) in Taiwan. In this study, we selected 12 050 patients diagnosed with PLA as our study cohort and 60 250 non-PLA patients as our comparison cohort. We individually tracked each subject for a 1-year period beginning with their index date to identify those who were subsequently diagnosed with any of the following infections: pneumonia, endophthalmitis, septic pulmonary embolism, pulmonary abscess, pleural empyema, meningitis, abscess of prostate, renal and perinephric abscess, epidural spinal abscess, osteomyelitis, necrotizing fasciitis, splenic abscess, psoas abscess and infectious endocarditis. We found that during the 1-year follow-up period, the subjects with PLA had a consistently higher incidence of all types of infections than comparison subjects. In particular, compared with subjects without PLA, the adjusted hazard ratios (HR) of pulmonary abscess, pleural empyema, renal and perinephric abscess, epidural spinal abscess and splenic abscess were 26.71, 18.56, 43.21, 51.32 and 126.51, respectively. We further analysed the HR of extra-hepatic Klebsiella pneumoniae infections among patients with PLA caused by K. pneumoniae. We found that the HR was higher for 12 of the 15 analysed extra-hepatic infections after restricting the analysis to only infections with K. pneumoniae aetiologies.

Introduction

Pyogenic liver abscess (PLA) is a potentially fatal disease. It currently accounts for three-quarters of all the liver abscesses occurring in industrialized countries [1], with one out of every 4500–7000 hospital admissions being due to a liver abscess [2,3]. Before the 1980s, the organism most commonly isolated from liver abscesses was Escherichia coli, but in Taiwan there has been a shift to Klebsiella pneumoniae, which is now highly endemic and the major cause of community-acquired pyogenic infections [4–7]. Some researchers are now suggesting that this same aetiological shift is also occurring in the USA [8], with K. pneumoniae more recently being found to be the primary causative agent there as well [9]. In addition to this aetiological shift, other factors affecting the prognosis of PLA include advancements in antibiotics and surgical techniques, which have greatly reduced mortality rates over the last two decades [10].

Although the treatment of PLA continues to evolve and is currently a topic of controversy among medical practitioners [11], surgical advancements have already reduced mortality rates to around 11.24% [12]. This can be compared with rates that were as high as 69% among PLA patients afflicted between 1928 and 1937 [12], and 24% in a study conducted on patients in the USA between 1966 and 1978 [13]. On account of the decreased mortality rates and increased number of PLA survivors, clinicians need to be more aware of the sequelae related to having suffered a PLA.

Patients with PLA often have high rates of bacteraemia and subsequent metastatic complications, and these rates are even higher when K. pneumoniae is the causative agent [4,14]. This is particularly disconcerting because severe complications including septic endogenous endophthalmitis, metastatic infections of the brain and lung, and necrotizing fasciitis have been reported in PLA patients all over the world [15–20]. Therefore, to better characterize the risk of sequelae to PLA, this study set out to provide risk estimates for a panel of infections subsequent to PLA using a population-based dataset in Taiwan.

Methods

Database

The data for this study were sourced from Taiwan’s National Health Insurance Research Database (NHIRD). The NHIRD is derived from the Taiwan National Health Insurance (NHI) programme and is maintained by the National Health Research Institute, Taipei, Taiwan. The NHIRD includes the registration files and original claims data for the reimbursement of 22.89 million of the country’s 22.96 million inhabitants as of 2008. Several studies have demonstrated the high validity of the data from the NHI programme [21–23].

Because the NHIRD consists of anonymous secondary data that are routinely released to the public for research purposes, this study was exempt from full review by the Taipei Medical University Institutional Review Board.

Study population

This retrospective cohort study included both a study cohort and a comparison cohort. In selecting our study cohort, we first identified 14 343 subjects from the NHIRD who visited outpatient care centres or were hospitalized with a principal diagnosis of PLA (International Classification of Diseases, 9th edition, clinical modification (ICD-9-CM) code 572.0) between 1 January 2006 and 31 December 2008. We assigned their first visits for PLA during this study period as their index dates. We then excluded subjects who had been diagnosed with PLA before their index date (= 1154) to include only new-onset patients and those <18 years old (= 148) to restrict our analysis to the adult population. We also excluded subjects who had received a diagnosis of any of the infections selected for investigation (including pneumonia, endophthalmitis, septic pulmonary embolism, pulmonary abscess, pleural empyema, meningitis, abscess of prostate, renal and perinephric abscess, epidural spinal abscess, osteomyelitis, pyogenic arthritis, necrotizing fasciitis, splenic abscess, psoas abscess and infectious endocarditis) within the 6 months before their index dates (n = 818). Ultimately, 12 050 subjects with PLA were included in the study cohort.

We also used the NHIRD to select the comparison cohort. In total, 60 250 subjects were randomly extracted from the NHIRD, with a ratio of 1:5, matched with the study cohort on age, sex, urbanization level (five levels, with one referring to the most urbanized and five referring to the least), and the year of index date (2006–2008). All 359 cities/towns in Taiwan were stratified into five groups ranked by urbanization level in accordance with a previous study with one referring to the ‘most urbanized’ and five referring to the ‘least urbanized’ [24]. We matched the distribution of urbanization level between the study subjects and the comparison subjects to help to assure that these two cohorts were reasonably similar in terms of unmeasured neighbourhood socio-economic characteristics. Their first use of medical services occurring in the index year was assigned as the index date. We also ensured that none of the selected comparison cohort had a history of PLA since the initiation of the Taiwan NHI programme and that none of the selected comparison subjects had been diagnosed with any type of the above-mentioned infections during the 6 months preceding their index dates.

As a result, we included 72 300 subjects in this study and individually tracked each subject for a 1-year period beginning with their index date to identify those who were subsequently diagnosed with any of the following infections: pneumonia (ICD-9-CM codes 480–486 or 487.0), endophthalmitis (ICD-9-CM codes 360.0–360.1), septic pulmonary embolism (ICD-9-CM codes 415.1), pulmonary abscess (ICD-9-CM codes 513.0), pleural empyema (ICD-9-CM codes 510), meningitis (ICD-9-CM codes 320–322), abscess of prostate (ICD-9-CM codes 601.2), renal and perinephric abscess (ICD-9-CM codes 590.2), epidural spinal abscess (ICD-9-CM codes 324.1), osteomyelitis (ICD-9-CM codes 730), pyogenic arthritis (ICD-9-CM codes 711.0), necrotizing fasciitis (ICD-9-CM codes 728.86), splenic abscess (ICD-9-CM codes 289.59), psoas abscess (ICD-9-CM codes 567.0) and infectious endocarditis (ICD-9-CM codes 421.0–421.1). These secondary infectious diseases were selected because they have been reported to be caused by extra-hepatic metastases from pyogenic liver abscesses in previous studies [25].

Klebsiella pneumoniae was identified by ICD-9-CM code 482.0 in the case of pneumonia, and ICD-9-CM code 041.3 in addition to a location code for other sources of infection.

Statistical analysis

We used the SAS statistical package for all statistical analyses. Pearson chi-square tests were used to examine the distributions of urbanization level, geographic region (northern, central, eastern and southern Taiwan) and monthly income and selected medical comorbidities (hypertension, diabetes, heart disease, lung disease, cancer and acute/chronic kidney disease) of both patients with and patients without PLA. These comorbidities were only included if they either occurred in an inpatient setting or appeared in two or more ambulatory care claims coded 6 months before the index date. Separate Stratified Cox proportional hazard regressions (stratified on age, sex, urbanization level and the year of index date) were conducted to calculate the hazard of each selected type of infection following a diagnosis of PLA. In addition, we further analysed the hazard of having suffered any type of infections and at least two types of infection between the study cohort and comparison cohort. A value of p <0.05 was considered statistically significant.

Results

Of the 72 300 sampled subjects, the mean age was 62.0 (±15.1) years and about 61% were male. We found that K. pneumoniae (ICD-9-CM code 041.3) was the causative organism in 73.0% of the patients with PLA in this study. Table 1 shows that patients with PLA were more likely than patients without PLA to have the following comorbidities after matching for age and sex: hypertension (p <0.001), diabetes (p <0.001), heart disease (p <0.001), cancer (p <0.001) and acute/chronic kidney disease (p <0.001). There was no significant difference in the prevalence of lung disease between patients with and without PLA (p 0.698). In addition, Table 1 shows that patients with PLA had a greater tendency to have no monthly income and to reside in the southern part of Taiwan than patients without PLA.

Table 1. Demographic characteristics and comorbid medical disorders for patients in Taiwan with pyogenic liver abscess and patients in the comparison cohort, 2006–2008 (n = 72 300)
VariablePatients with pyogenic liver abscess (= 12 050)Comparison patients (= 60 250)p value
Total no.Column %Total no.Column %
  1. aHeart disease: ischaemic heart disease, arrhythmias and heart failure.

  2. bLung disease: chronic obstructive pulmonary disease, asthma, pulmonary emphysema and pneumoconiosis.

Sex
 Male733260.836 66060.81.000
 Female471839.223 59039.2
Age (years)
 18–39126410.5632010.51.000
 40–49185815.4929015.4
 50–59292724.314 63524.3
 60–69256021.312 80021.3
 70–79247320.512 36520.5
 >799688.048408.0
Urbanization level
 1 (most urbanized)337428.016 87028.01.000
 2324226.916 21026.9
 3194216.1971016.1
 4189415.7947015.7
 (least urbanized)159813.3799013.3
Hypertension605550.326 31443.7<0.001
Diabetes532244.212 16020.2<0.001
Heart diseasea249120.710 87018.0<0.001
Lung diseaseb10448.752868.80.698
Acute/chronic kidney disease134111.133935.6<0.001
Cancer324126.933105.4<0.001
Monthly income
 NT$0–15 840549245.621 19635.2<0.001
 NT$15 841–25 000427135.425 87543.0
 ≥NT$25 001228719.013 17921.8
Geographic region
 Northern528943.928 02846.5<0.001
 Central286723.814 49324.1
 Southern357629.716 16226.8
 Eastern3182.615672.6

The incidence of infection during the 1-year follow-up period for these two cohorts is presented in Table 2. Subjects with PLA had a consistently higher incidence of all types of infection than subjects without PLA during the 1-year follow-up period. Table 2 also presents the crude and adjusted hazard ratio (HR) for different types of infection between subjects with and without PLA. After adjusting for patient monthly income, geographic location, hypertension, diabetes, heart disease, lung disease, cancer (p <0.001) and acute/chronic kidney disease, stratified Cox proportional hazards regressions (stratified on age, sex, urbanization level and index year) consistently revealed that when compared with subjects without PLA, subjects with PLA had higher HRs for all types of infection during the 1-year follow-up period. The risks of some of the selected infections were particularly high. For example, compared with subjects without PLA, the adjusted HR of pulmonary abscess, pleural empyema, renal and perinephric abscess, epidural spinal abscess and splenic abscess were 26.71, 18.56, 43.21, 51.32 and 126.51, respectively. Fig. 1 presents the pneumonia-free survival curves by the Kaplan–Meier method.

Table 2. Crude and adjusted hazard ratios for subsequent extra-hepatic infections among the sample patients during the 365-day follow-up period starting from the index date (n = 72 300)
Types of subsequent extra-hepatic infectionsPatients with PLA (= 12 050)Comparison patients (= 60 250)Patients with pyogenic liver abscess versus comparison
No. (%)No. (%)Crude HR, 95% CIAdjusted HRb, 95% CI
  1. 95% CI, 95% confidence interval; HR, hazard ratio; PLA, pyogenic liver abscess.

  2. aIndicates p <0.001; hazard ratio was calculated by the stratified Cox regression model (stratified on age, sex, urbanization level, and index date).

  3. bAdjustments are made for patient monthly income, geographic location, hypertension, diabetes, heart disease, lung disease, cancer and acute/chronic kidney disease.

Pneumonia1516 (12.6)2405 (4.0)3.46 (3.24–3.70)a3.25 (3.00–3.52)a
Endophthalmitis110 (0.9)33 (0.1)16.81 (11.39–24.82)a16.53 (10.99–24.85)a
Septic pulmonary embolism54 (0.5)23 (0.1)11.79 (7.23–19.21)a9.52 (5.63–16.08)a
Pulmonary abscess91 (0.8)14 (0.1)32.7 (18.65–57.49)a26.71 (14.86–47.99)a
Pleural empyema154 (1.3)33 (0.1)23.62 (16.21–34.42)a18.56 (12.48–27.62)a
Meningitis70 (0.6)23 (0.1)15.30 (9.55–24.52)a14.19 (8.64–23.29)a
Abscess of prostate16 (0.1)2 (0.1)40.05 (9.21–174.21)a18.31 (4.02–83.40)a
Renal and perinephric abscess59 (0.5)6 (0.1)49.40 (21.33–114.44)a43.21 (18.25–102.27)a
Epidural spinal abscess12 (0.1)1 (0.1)60.05 (7.81–461.86)a51.32 (6.34–415.64)a
Osteomyelitis92 (0.8)117 (0.2)3.96 (3.01–5.20)a3.06 (2.27–4.12)a
Pyogenic arthritis37 (0.3)32 (0.1)5.80 (3.61–9.31)a4.71 (2.83–7.84)a
Necrotizing fasciitis46 (0.4)37 (0.1)6.24 (4.04–9.62)a3.83 (2.40–6.12)a
Splenic abscess54 (0.5)2 (0.1)135.60 (33.06–556.23)a126.51 (30.37–527.09)a
Psoas abscess16 (0.1)4 (0.1)20.03a (6.69–59.91)21.94a (6.96–69.13)
Infectious endocarditis34 (0.3)10 (0.1)17.05 (8.42–34.51)a16.20 (7.72–34.00)a
Figure 1.

 Pneumonia-free survival rates for patients with pyogenic liver abscess and comparison patients in Taiwan.

Table 3 further analysed the incidence and HR of extra-hepatic K. pneumoniae infections among patients with PLA caused by K. pneumoniae. We only analysed for K. pneumoniae because it was the most prevalent infectious agent in this study.

Table 3. Crude and adjusted hazard ratios for extra-hepatic Klebsiella pneumoniae infections among the sample patients during the 365-day follow-up period starting from the index date (n = 70 089)
Types of extra-hepatic Klebsiella pneumoniae infectionsPatients with K. pneumoniae-caused PLA (= 9839)Comparison (= 60 250)Patients with pyogenic liver abscess versus Comparison
No. (%)No. (%)Adjusted HRb, 95% CI
  1. 95% CI, 95% confidence interval; HR, hazard ratio; PLA, pyogenic liver abscess.

  2. ap <0.001, bp <0.01. Hazard ratio was calculated by the stratified Cox regression model (stratified on age, sex, urbanization level and index date).

  3. bAdjustments are made for patient monthly income, geographic location, hypertension, diabetes, heart disease, lung disease, cancer and acute/chronic kidney disease.

Pneumonia1021 (10.4)913 (1.5)6.96 (6.33–7.64) a
Endophthalmitis84 (0.9)33 (0.1)15.59 (10.20–23.83)a
Septic pulmonary embolism44 (0.5)19 (0.1)11.92 (6.70–21.22)a
Pulmonary abscess69 (0.7)8 (0.1)42.81 (20.07–91.32)a
Pleural empyema121 (1.2)24 (0.1)24.94 (15.73–39.54)a
Meningitis51 (0.5)15 (0.1)18.83 (10.28–34.47)a
Abscess of prostate11 (0.1)1 (0.1)32.91 (4.09–264.69)b
Renal and perinephric abscess45 (0.5)5 (0.1)51.45 (19.95–132.69)a
Epidural spinal abscess8 (0.1)
Osteomyelitis63 (0.6)94 (0.2)3.41 (2.42–4.82)a
Pyogenic arthritis28 (0.3)24 (0.1)5.67 (3.15–10.20)a
Necrotizing fasciitis34 (0.4)24 (0.1)5.41 (3.08–9.50)a
Splenic abscess42 (0.4)2 (0.1)113.15 (26.80–477.81)a
Psoas abscess12 (0.1)3 (0.1)24.98a (6.62–94.24)
Infectious endocarditis24 (0.2)3 (0.1)47.53 (13.88–162.71)a

Discussion

This study reported risk estimates for a panel of infections subsequent to suffering a PLA, which ranged between 3.06 for osteomyelitis, and 126.51 for splenic abscess. We also conducted another analysis on the incidence and HR of extra-hepatic K. pneumoniae infections among patients with PLA caused by K. pneumoniae to gain better insight into the possibility that the mechanism underlying the increased risks detected in this study involved metastasis. We found that the HR for 12 of the 15 analysed extra-hepatic infections increased after restricting the analysis to only infections with K. pneumoniae aetiologies.

The gruop of PLA survivors afflicted with subsequent infections is likely to grow based on the aetiological shift among PLA patients. Before the 1980s, the most commonly isolated organisms from liver abscesses were E. coli, but in Taiwan there has been a shift to K. pneumoniae as the principal causative organism, which is now not only endemic but also the major cause of all community-acquired pyogenic infections [4–7]. This high prevalence of PLA cases aetiologically characterized by K. pneumoniae in this study supports the findings of a previous study conducted on PLA patients between 1990 and 1996, which found 87.9% of all the PLA cases in one veteran hospital in Taiwan to be caused by K. pneumoniae alone, with the remainder being caused by mixed flora [26].

Hence, it is likely that the increased risks for infection following PLA observed in this study stem from factors accompanying the aetiological shift to K. pneumoniae as the primary causative agent among PLA infections. Two such accompanying factors through which this mechanism may proceed include the increased risks for bacteraemia and metastatic complications among K. pneumoniae-caused PLA. In one study conducted on all patients in the USA with PLA caused by K. pneumoniae, it was found that 83% had concurrent bacteraemia [4]. This can be compared with another US study conducted on records taken from the period between 1994 and 2005, which only reported rates of bacteraemia reaching 52.1% (95% CI = 50.9–53.3) among 8286 PLA patients with the most common bacterial sources identified as streptococcus (29.5%) and E. coli (18.1%) [14].

In addition to increased rates of bacteraemia, the same study found that 28% of the patients with PLA arising from K. pneumoniae also suffered from metastatic complications [4]. This is supported by a study conducted in Taiwan on 225 PLA cases sourced from between 1995 and 2000, which found PLA patients infected with K. pneumoniae to be more likely to have had metastatic complications (adjusted OR = 5.0; 95% CI = 1.1–47) than cases caused by other bacteria after controlling for age, sex and the duration of symptoms before admission [25]. The same study also found cases presenting with bacteraemia, which is more likely among patients with PLA caused by K. pneumoniae [4,14], to be at an increased risk for metastatic complications (adjusted OR = 5.4; 95% CI = 1.4–30) [25], suggesting that patients with PLA caused by K. pneumoniae are particularly vulnerable to metastasis.

Therefore, it is possible that the increased risks for extra-hepatic infection following PLA may at least in part be explained through a metastatic mechanism. This is supported by our further analysis on the incidence and HR of extra-hepatic K. pneumoniae infections among patients with PLA caused by K. pneumoniae (Table 2). We observed that the HR was higher for 12 of the 15 analysed extra-hepatic infections after restricting the analysis to only infections with K. pneumoniae aetiologies.

This study suffered from two principal limitations. First, some clinically relevant patient and lifestyle information, such as smoking status, alcohol consumption and dietary habits, body mass index, and pharmaceutical use was not available through the administrative dataset. Another study recently conducted in Taiwan has suggested that diabetes and alcoholism are significant risk factors for developing extra-hepatic metastatic infections from PLA [25], so in addition to adjusting for urbanization level, which may reflect lifestyle, we also adjusted for both obesity and alcohol abuse/dependence syndrome. Nevertheless, the association between PLA and subsequent infections detected in this study may be partially explained by the residual confounding of these unadjusted factors.

Second, the risk of subsequent infection may vary with the severity of the PLA infection; however, there were no laboratory measures that could serve as a proxy for severity (antibody titres, bacterial concentrations, bacterial identification, or other biomarkers) available in the dataset.

This study suggested an increased risk of infection among individuals with a PLA. Future studies will be needed to elucidate the mechanisms by which PLA is associated with subsequent infections, and to clarify whether the identity of the causative organism influences the risk of subsequent infection.

Acknowledgments

This study was supported by a grant from the 101CGH-TMU-01-1. This study is based in part on data from the NationalHealth Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health, Taiwan and managed by the National Health Research Institutes. The interpretations and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health, or the National Health Research Institutes.

Transparency Declaration

The authors have no proprietary or commercial interest in any materials mentioned in this article.

Ancillary