2023 Volume 46 Issue 2 Pages 237-244
Community-acquired pneumonia (CAP) is an acute pulmonary parenchymal infection acquired outside the hospital. The utility of blood cultures in inpatients with CAP to reduce mortality and length of hospital stay is controversial. This study aimed to determine the utility of blood cultures on the first day of hospitalization for CAP inpatients and its influence on mortality, length of hospital stay, and antibiotics use. We conducted a fact-finding survey on the implementation of blood culture in inpatients with CAP in Japan. A propensity score (PS)-matched analysis based on the National Database of Health Insurance Claims and Specific Health Check-ups of Japan database was conducted. Overall, 163173 patients were included in the analysis, and PS matching extracted 68104 pairs. The results of the comparison between the PS-matched blood culture group and PS-matched control group were as follows: mortality and length of hospital stay were significantly lower in the PS-matched blood culture group than in the control group. The adjusted odds ratio (OR) (95% confidence interval (CI)) for in-hospital mortality with blood culture test was 0.73 (0.68–0.79). Moreover, for days of antibiotic usage, number of antibiotics used were significantly higher in the PS-matched blood culture group than that in the control group. Our findings indicated that performing a blood culture on the first day of hospitalization for inpatients with CAP was associated with reduced mortality. To our knowledge, this is the largest epidemiological study to assess the utility of blood culture in Japanese inpatients with CAP. This testing method shows potential for application in clinical practice.
Pneumonia is a pulmonary parenchymal infection and is categorized as community-acquired pneumonia (CAP), healthcare-associated pneumonia, and nursing and healthcare-associated pneumonia in Japan.1) CAP is an acute pulmonary parenchymal infection acquired outside the hospital. It is one of the most common and morbid conditions encountered in clinical practice.2)
The diagnosis of CAP is generally based on the presence of an infiltrate on chest imaging in a patient with a clinically compatible symptoms (e.g., fever, dyspnea, cough, and sputum production).3) For patients with a clinical diagnosis of CAP, the next steps in management are defining the severity of illness and determining the most appropriate place for treatment and care. Determining the severity of illness is based on clinical judgment and can be supplemented by the use of severity scores. The most commonly used severity score in Japan is the A-DROP score. The A-DROP score is calculated for each of the five states: A (Age: male: ≥ 70 y; female: ≥ 75 y), D (Dehydration: blood urea nitrogen ≥ 21 mg/dL or dehydration (+)), R (Respiration: oxyhemoglobin saturation measured by pulse oximetry ≤90%), O (Orientation: altered consciousness (+)), and P (systolic blood pressure: ≤ 90 mmHg); each has a score of 1 point, and a score of ≥ 4 is defined as severe.1)
In addition to A-DROP, the presence or absence of sepsis was added to assess the severity of CAP and need for hospitalization and admission to the intensive care unit (ICU).1) CAP is the second most common cause of hospitalization and the most common cause of death due to infection.4) Mortality can be attributed to CAP either directly (e.g., overwhelming sepsis or respiratory failure) or indirectly due to cardiovascular events or other comorbid complications (e.g., advanced chronic obstructive pulmonary disease). The mortality rate directly attributable to CAP is approximately 6%.5)
Sepsis is defined as a life-threatening organ dysfunction caused by a dysregulated host response to infection.6) Severe sepsis and septic shock develop in approximately 48 and 5% of patients hospitalized with CAP, respectively.7) Sepsis is associated with a high mortality rate, and the reported rates depend on the method of data collection, with the estimated range being 10–52%.8) In Japan, the quick sequential organ failure assessment (qSOFA)6) is recommended as a criterion for sepsis in patients with CAP.1) The qSOFA was proposed as a simple tool for screening sepsis outside the ICU. According to the qSOFA criteria, sepsis is diagnosed if two or more of the following criteria are satisfied: alteration in mental status (Glasgow Coma Scale9) < 15), systolic blood pressure < 100 mmHg, respiratory rate ≥ 22/min.6) In contrast, bacteremia is defined as presence of viable bacteria in the blood.8) Jones and Lowes10) reported that, in a study of 270 blood cultures, 95% of positive blood cultures were associated with sepsis or septic shock. However, sepsis can also occur outside of infections, and there are no clear guidelines to help clinicians identify the presence of infection or to causally link an identified organism with sepsis.8) Blood cultures help clinicians in identifying complications (e.g., bacteremia) or making alternate diagnoses (e.g., endocarditis), which might otherwise have remained undetected. The major disadvantages of obtaining blood cultures include their low diagnostic yield and the low degree of certainty that their results would improve clinical outcomes. In a retrospective cohort study in the United States that evaluated more than 130000 patients with pneumonia, blood cultures were positive in 4.7% of patients.11) While the degree to which culture results lead to changes in management that improve outcomes has not been well studied, available data suggest that the number might be low.12,13)
For these reasons, the importance of performing blood cultures for all hospitalized CAP patients is debated.3) In the 2007 American Thoracic Society and Infectious Diseases Society of America (ATS/IDSA) guideline,2) blood culture was primarily recommended for patients with severe disease. The 2019 ATS/IDSA guideline3) suggests that obtaining blood cultures for adults with CAP managed in the hospital setting should not be routine (conditional recommendation, very low quality of evidence). In the same guideline, blood cultures were recommended for patients with severe disease and for all inpatients empirically treated for methicillin-resistant Staphylococcus aureus or Pseudomonas aeruginosa. However, the quality of evidence is very low.3) Similarly, the 2017 Japanese guidelines state that although the evidence is not high, blood cultures are necessary for detecting severe pneumonia and are recommended for educational purposes.1) Therefore, more research is needed to evaluate the utility of blood cultures in patients with CAP. However, as the purpose of microbiological testing of patients with CAP is to obtain additional diagnostic information that could improve the quality of treatment decisions, it is ethically challenging to randomly assign patients to the non-implementation group.
Therefore, we conducted a fact-finding survey on the implementation of blood cultures for CAP inpatients in Japan requiring hospitalization using the National Database of Health Insurance Claims and Specific Health Check-ups of Japan (NDB), which contains more than 90% of Japanese medical billing information.
The present study aimed to determine the utility of blood cultures on the first day of hospitalization for CAP inpatients and its effects on mortality, length of hospital stay, and antibiotic use.
A propensity score (PS)-matched analysis based on a claims database was conducted.
Ethics ApprovalThis study was approved by the Nagasaki International University Ethics Committee (Approval No. 44, 18 June 2020). Since the data utilized in this study were already anonymized by the database provider, the study was exempted from the requirement to obtain informed consent from individual patients according to the local ethical guidelines for epidemiological research.
Data SourceThe NDB was used as the data source. The NDB, one of the largest health-related databases worldwide (more than 100 million IDs, comparable to the total population in Japan), contains all electronic claims data reimbursed by Japan’s public medical insurance system, excluding public assistance. As of 2017, more than 98% of claims have been issued electronically.14) This database includes information on the following variables: patient data, admission and discharge status, diagnoses, and the drugs and procedures used.
Japan’s insurance claim method has traditionally involved payment for medical treatment. However, to reduce the variation among medical institutions, the Diagnosis Procedure Combination/Per-Diem Payment System (DPC), which is a flat-rate calculation method based on diagnostic group classification, has been introduced. Hospitals must meet national standards to use the DPC, and the number of hospitals using the DPC has gradually increased (83% of acute care beds in 2018).
The DPC data were included in the NDB. Since 2016, the A-DROP score has been incorporated into the DPC coding for pneumonia. These data are provided to researchers after removing personal information and irreversibly anonymizing the data.
Data AvailabilityData used in the present study cannot be shared publicly because the data include personal information. Data are available from the Ministry of Health, Labour, and Welfare of Japan and will be provided to researchers who meet the criteria for access to confidential data. Contact information is below: E-mail teikyo_rezept@kits.nttdata.co.jp.
Study PopulationData on patients with admission date of April 1, 2016, or later, and a discharge date of March 31, 2017, or earlier, were collected from the NDB. The Japanese guidelines 2017 state that although the evidence is not of high quality, blood cultures are necessary for severe pneumonia and are recommended for educational purposes.1) Therefore, the study population was selected to eliminate the influence of these guidelines. Patients aged >15 years with CAP but without surgery were identified according to the DPC code at discharge, and their data were extracted.
In addition, the following data were excluded: data other than DPC received in the same month (e.g., mismatch of DPC code between admission and discharge, and duplicate DPC code) (simple error); presence of non-DPC billing (credibility issues); outcomes such as “outside death,” “others” (credibility issues); no antibiotics used (credibility issues); scheduled hospitalization (possible non-CAP, possible presence of prior treatment); short-term returning home during hospitalization (affects hospitalization period); another newly detected infectious disease during hospitalization (coexistence of other treatment); and prescription of antiviral drugs (possibility of prophylactic administration such as among patients with human immunodeficiency virus infection).
Data CollectionThe following data were collected: blood culture performed or not performed on the day of admission, patient age, sex, A-DROP-score, outcome, secondary disease, and blood culture required or not required according to the ATS/IDSA Guideline 20072) (hereinafter, cautionary diseases: e.g., pancytopenia, alcohol dependence, liver disease). Additionally, data were collected on splenia, urinary pneumococcal antigen (substitute for measurement), duration of antibiotic use, types of antibiotics, presence or absence of ICU admission, and use of a ventilator.
Primary and Secondary OutcomesThe primary outcome was the in-hospital mortality rate, and the secondary outcomes included ICU admission, length of ICU stay, ventilator use, number of days of ventilator use, length of hospitalization, number of days with antibiotics use, and number of antibiotics used.
Data AnalysisThe characteristics of patients included in this study were summarized using frequencies and percentages for categorical data, and medians and interquartile ranges (IQRs) for continuous data.
We used PS matching to reduce the potential baseline differences between the blood culture and control groups.15) The PS was defined as the probability that a patient would receive a blood culture. First, the PS was estimated in a binary logistic regression model with implementation of blood culture as the outcome. Some PS matching of incorporating different sets of covariates were conducted with 1 : 1 nearest neighbor matching using a caliper method (caliper: 0.20). In each analysis, the model set that yielded the best matched cohort was identified based on the most balanced distribution of PS and the best balance in individual covariates between the two groups. The model fit was evaluated using the concordance index (C-index). The statistical balance between the two groups was evaluated. We used a standardized difference (Std diff.) to measure covariate balance, whereby an absolute Std. diff. of above 10% (Std diff. < 0.1) represented meaningful imbalance.16) Finally, PS were calculated by four covariates (sex, age [≥65 y], A-DROP score [severe: 4 or 5], and secondary disease) as the predictors.
We compared the PS-matched blood culture group with the PS-matched control group using Wilcoxon’s rank-sum test for continuous variables, and Fisher’s exact test was performed for categorical variables.
A multivariable logistic regression model yielding odds ratios (ORs) and 95% confidence intervals (95% CIs) was used to identify factors that may be associated with in-hospital mortality. p < 0.05 was considered to indicate statistical significance.
The expect value (E-value) is used for sensitivity analysis.17) The E-value is one calculated as the minimum influence of the unmeasured confounding factor when there is an unmeasured confounding factor having sufficient influence to overturn the statistical result. This unmeasured confounding factor is defined as influencing both cause and effect.18) We calculated an E-value that represents the magnitude of the effect of unmeasured confounders on both blood culture test and in-hospital mortality.
All data analyses were performed using JMP Pro 16 (SAS Institute Inc., Cary, NC, U.S.A.), with a significance level of 5%. Data cleaning was performed using Visual Mining Studio 8.6 (NTT DATA Mathematical Systems Inc., Tokyo, Japan).
A total of 163173 patients were included in the study, of whom 42.9% (69990 patients) underwent blood culture on the first day of hospitalization (Fig. 1, Table 1). The characteristics of patients included in this study are illustrated in Table 1. The in-hospital mortality (outcome: death) was 5.6%. Proportion of patients aged >65 years was 79.9%. The proportion of patients with A-DROP score (severe: 4 or 5) was 5.3%. The median of length of hospital stay was 11 d. ICU admission occurred in 0.7% of patients. The median of number of days with antibiotics use was 8, and the median of number of antibiotics used was 2.
The flow diagram illustrates the inclusion and exclusion criteria for the study population. DPC, The Diagnosis Procedure Combination/Per-Diem Payment System.
| Characteristics | n = 163173 | |
|---|---|---|
| n | (%) | |
| Blood culture test | 69990 | (42.9) |
| Sex (male) | 93925 | (57.6) |
| Outcome | ||
| Complete cure | 13687 | (8.4) |
| Cure | 137075 | (84.0) |
| Remission | 406 | (0.2) |
| Invariant | 2662 | (1.6) |
| Exacerbation | 236 | (0.1) |
| Death | 9107 | (5.6) |
| Age, years | ||
| 15–19 | 3423 | (2.1) |
| 20–24 | 2450 | (1.5) |
| 25–29 | 2284 | (1.4) |
| 30–34 | 2815 | (1.7) |
| 35–39 | 2945 | (1.8) |
| 40–44 | 2926 | (1.8) |
| 45–49 | 2801 | (1.7) |
| 50–54 | 2757 | (1.7) |
| 55–59 | 3824 | (2.3) |
| 60–64 | 6532 | (4.0) |
| 65–69 | 12513 | (7.7) |
| 70–74 | 15686 | (9.6) |
| 75–79 | 22070 | (13.5) |
| 80–84 | 28371 | (17.4) |
| 85 over | 51776 | (31.7) |
| Over 65 years | 130416 | (79.9) |
| Severity classification: A-DROP score | ||
| 0 | 29344 | (18.0) |
| 1 | 51801 | (31.7) |
| 2 | 48579 | (29.8) |
| 3 | 24873 | (15.2) |
| 4 | 7026 | (4.3) |
| 5 | 1550 | (0.9) |
| Severe pneumonia: A-DROP (4 or 5) | 8576 | (5.3) |
| Secondary disease | 40166 | (24.6) |
| Heartfailure I50 | 29220 | (17.9) |
| Pleural effusion, not elsewhere classified J90x | 10137 | (6.2) |
| Enterocolitis due to Clostridium difficile A047 | 369 | (0.2) |
| Pleural plaque without asbestos J929 | 129 | (0.1) |
| Pleural plaque with presence of asbestos J920 | 34 | (0.0) |
| Other specified pleural conditions J948 | 28 | (0.0) |
| Others | 249 | (0.2) |
| Cautionary diseases | 5403 | (3.3) |
| ICU admission | 1134 | (0.7) |
| Length of ICU stay, median (IQR), days | 0 | (0–0) |
| Ventilator use | 5618 | (3.4) |
| Number of days of ventilator use, median (IQR), days | 0 | (0–0) |
| Length of hospital stay, median (IQR), days | 11 | (8–16) |
| Number of days with antibiotics use, median (IQR), days | 8 | (6–12) |
| Number of antibiotics used, median (IQR), types | 2 | (1–2) |
Cautionary diseases: the disease required a blood culture according to the ATS/IDSA Guideline 2007. Abbreviations: IQR, interquartile range. Data are expressed as numbers (n); (%) unless otherwise indicated. The ICD10 code (International Classification of Diseases 10th Revision) of the disease is mentioned on the right. ICD10 code of the disease is mentioned on the right.
PS matching extracted 68104 pairs. Table 2 presents the details of the cohorts created by PS matching. Before PS matching, the proportion of males, patients aged >65 years, patients with severe pneumonia (A-DROP score: 4 or 5), and patients with an ICU admission in the blood culture group were significantly higher than in the control group. After PS matching, because a Std diff. < 0.1 was obtained for all variables, the covariate balance in the matched cohort was considerably improved. The C-index of this model was 0.56129.
| Characteristics | Blood culture group | Control group | p-Value | Std diff. | Propensity-matched Blood culture group | Propensity-matched Control group | p-Value | Std diff. | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| n = 69990 | n = 93183 | n = 68104 | n = 68104 | ||||||||||
| Sex (male) | 41921 | (59.9) | 52004 | (55.8) | <0.0001 | 0.0831 | 40270 | 59.1 | 40255 | 59.1 | 0.9385 | 0.0004 | |
| Age, years | 15–19 | 1200 | (1.7) | 2223 | (2.4) | <0.0001 | 0.0494 | 1194 | 1.8 | 1378 | 2.0 | 0.0003 | 0.0199 |
| 20–24 | 965 | (1.4) | 1485 | (1.6) | 0.0004 | 0.0165 | 959 | 1.4 | 955 | 1.4 | 0.9449 | 0.0005 | |
| 25–29 | 879 | (1.3) | 1405 | (1.5) | <0.0001 | 0.0170 | 864 | 1.3 | 915 | 1.3 | 0.2328 | 0.0066 | |
| 30–34 | 1110 | (1.6) | 1705 | (1.8) | 0.0002 | 0.0155 | 1093 | 1.6 | 1182 | 1.7 | 0.0628 | 0.0102 | |
| 35–39 | 1186 | (1.7) | 1759 | (1.9) | 0.0038 | 0.0150 | 1163 | 1.7 | 1207 | 1.8 | 0.3729 | 0.0049 | |
| 40–44 | 1193 | (1.7) | 1733 | (1.9) | 0.0194 | 0.0150 | 1174 | 1.7 | 1194 | 1.8 | 0.6937 | 0.0022 | |
| 45–49 | 1227 | (1.8) | 1574 | (1.7) | 0.3262 | 0.0076 | 1196 | 1.8 | 1109 | 1.6 | 0.0708 | 0.0075 | |
| 50–54 | 1209 | (1.7) | 1548 | (1.7) | 0.3129 | 0.0000 | 1167 | 1.7 | 1102 | 1.6 | 0.1754 | 0.0075 | |
| 55–59 | 1718 | (2.5) | 2106 | (2.3) | 0.0104 | 0.0131 | 1671 | 2.5 | 1492 | 2.2 | 0.0014 | 0.0175 | |
| 60–64 | 2820 | (4.0) | 3712 | (4.0) | 0.646 | 0.0000 | 2737 | 4.0 | 2761 | 4.1 | 0.7515 | 0.0018 | |
| 65–69 | 5621 | (8.0) | 6892 | (7.4) | <0.0001 | 0.0225 | 5440 | 8.0 | 5116 | 7.5 | 0.0011 | 0.0178 | |
| 70–74 | 6993 | (10.0) | 8693 | (9.3) | <0.0001 | 0.0237 | 6759 | 9.9 | 6519 | 9.6 | 0.0290 | 0.0119 | |
| 75–79 | 9954 | (14.2) | 12116 | (13.0) | <0.0001 | 0.0350 | 9650 | 14.2 | 9055 | 13.3 | <0.0001 | 0.0254 | |
| 80–84 | 12342 | (17.6) | 16029 | (17.2) | 0.0228 | 0.0106 | 12026 | 17.7 | 11873 | 17.4 | 0.2789 | 0.0059 | |
| 85 over | 21573 | (30.8) | 30203 | (32.4) | <0.0001 | 0.0344 | 21011 | 30.9 | 22246 | 32.7 | <0.0001 | 0.0390 | |
| Age (over 65 years) | 56483 | (80.7) | 73933 | (79.3) | <0.0001 | 0.0350 | 54886 | 80.6 | 54809 | 80.5 | 0.6030 | 0.0029 | |
| Severity classification: A-DROP score | 0 | 11078 | (15.8) | 18266 | (19.6) | <0.0001 | 0.0997 | 10876 | 16.0 | 12576 | 18.5 | <0.0001 | 0.0662 |
| 1 | 20856 | (29.8) | 30945 | (33.2) | <0.0001 | 0.0732 | 20478 | 30.1 | 22334 | 32.8 | <0.0001 | 0.0587 | |
| 2 | 21459 | (30.7) | 27120 | (29.1) | <0.0001 | 0.0350 | 21065 | 30.9 | 19912 | 29.2 | <0.0001 | 0.0369 | |
| 3 | 12097 | (17.3) | 12776 | (13.7) | <0.0001 | 0.0996 | 11849 | 17.4 | 9461 | 13.9 | <0.0001 | 0.0966 | |
| 4 | 3682 | (5.3) | 3344 | (3.6) | <0.0001 | 0.0402 | 3136 | 4.6 | 3129 | 4.6 | 0.9381 | 0.0005 | |
| 5 | 818 | (1.2) | 732 | (0.8) | <0.0001 | 0.0402 | 700 | 1.0 | 692 | 1.0 | 0.8504 | 0.0012 | |
| A-DROP score (severe: 4 or 5) | 4500 | (6.4) | 4076 | (4.4) | <0.0001 | 0.0886 | 3836 | 5.6 | 3821 | 5.6 | 0.8692 | 0.0010 | |
| Secondary disease | |||||||||||||
| Enterocolitis due to Clostridium difficile | A047 | 205 | 0.3 | 164 | 0.2 | <0.0001 | 0.0242 | 198 | 0.3 | 115 | 0.2 | <0.0001 | 0.0255 |
| Heart failure | I50 | 12139 | 17.3 | 17081 | 18.3 | <0.0001 | 0.0258 | 11843 | 17.4 | 12595 | 18.5 | <0.0001 | 0.0288 |
| Pleural effusion, not elsewhere classified | J90x | 4395 | 6.3 | 5742 | 6.2 | 0.336 | 0.0049 | 4238 | 6.2 | 4330 | 6.4 | 0.3098 | 0.0056 |
| Pleural plaque with presence of asbestos | J920 | 14 | 0.0 | 20 | 0.0 | 0.8397 | 0.0010 | 13 | 0.0 | 16 | 0.0 | 0.7103 | 0.0030 |
| Pleural plaque without asbestos | J929 | 55 | 0.1 | 74 | 0.1 | 0.9762 | 0.0003 | 54 | 0.1 | 56 | 0.1 | 0.9240 | 0.0010 |
| Fibrothorax | J941 | *** | 0.0 | *** | 0.0 | 0.9609 | 0.0016 | *** | 0.0 | *** | 0.0 | 1.0000 | 0.0024 |
| Other specified pleural conditions | J948 | *** | 0.0 | 27 | 0.0 | <0.0001 | 0.0223 | *** | 0.0 | 22 | 0.0 | <0.0001 | 0.0237 |
| Pleural condition, unspecified | J949 | *** | 0.0 | *** | 0.0 | 0.8082 | 0.0005 | *** | 0.0 | *** | 0.0 | 0.7237 | 0.0000 |
| Chronic respiratory failure | J961 | *** | 0.0 | *** | 0.0 | 0.8036 | 0.0040 | *** | 0.0 | *** | 0.0 | 0.4795 | 0.0000 |
| Other diseases of liver | K76 | *** | 0.0 | *** | 0.0 | 0.7481 | 0.0007 | *** | 0.0 | *** | 0.0 | 1.0000 | 0.0024 |
| Other symptoms and signs involving the circulatory and respiratory system | R09 | *** | 0.0 | *** | 0.0 | 0.8036 | 0.0017 | *** | 0.0 | *** | 0.0 | 1.0000 | 0.0031 |
| Coded as Unknown | YYY | 81 | 0.1 | 129 | 0.1 | 0.2315 | 0.0064 | 79 | 0.1 | 91 | 0.1 | 0.3986 | 0.0050 |
| Cautionary diseases | 2313 | (3.3) | 3090 | (3.3) | 0.911 | 0.0000 | 2225 | 3.3 | 2260 | 3.3 | 0.6057 | 0.0029 | |
Cautionary diseases: the disease required a blood culture according to the ATS/IDSA Guideline 2007. Abbreviations: Std. diff., standardized difference. Data are expressed as numbers (n); (%). Unless otherwise indicated. The ICD10 code (International Classification of Diseases 10th Revision) of the disease is mentioned on the right. Fisher’s exact test was performed for categorical variables. Chi-square test was performed for ordinal variables. ***, According to the regulations of the Ministry of Health, Labour and Welfare, which is the source of information, total numbers less than 10 were masked.
Table 3 shows a comparison of the blood culture and control groups after PS matching. The mortality in the blood culture group was significantly lower than in the control group (5.2 vs. 6.3%, p < 0.0001). Patients in the blood culture group also had significantly shorter hospital stay than patients in the control group (average: 13.58 d vs. average: 13.69 d, p < 0.0001). However, the number of days with antibiotics use and number of antibiotics used were significantly higher in the blood culture group. Moreover, ICU admission and ventilator use were significantly higher in the blood culture than in the control group.
| Outcome | Propensity-matched blood culture group | Propensity-matched control group | p-Value | ||||
|---|---|---|---|---|---|---|---|
| n = 68104 | n = 68104 | ||||||
| Average | Median | % Or range | Average | Median | % Or range | ||
| Primary outcome | |||||||
| Death, No. (%) | 3557 | (5.2) | 4258 | (6.3) | <0.0001 | ||
| Secondary outcome | |||||||
| ICU admission, No. (%) | 669 | (0.98) | 327 | (0.48) | <0.0001 | ||
| Length of ICU stay, average, median (range), days | 0.04 | 0 | (0–14) | 0.02 | 0 | (0–14) | <0.0001 |
| Ventilator use, No. (%) | 2740 | (4.0) | 2118 | (3.1) | <0.0001 | ||
| Number of days of ventilator use, average, median (range), days | 0.28 | 0 | (0–60) | 0.24 | 0 | (0–60) | <0.0001 |
| Length of hospital stay, average, median (range), days | 13.58 | 11 | (1–60) | 13.69 | 11 | (1–60) | 0.0358 |
| Number of days with antibiotics use, average, median (range), days | 10.38 | 9 | (1–60) | 10.11 | 8 | (1–60) | <0.0001 |
| Number of antibiotics used, average, median (range), types | 2.01 | 2 | (1–10) | 1.88 | 2 | (1–12) | <0.0001 |
Abbreviations: Std. Diff., standardized difference; ICU, intensive care unit. Wilcoxon’s rank-sum test was performed for continuous variables, and Fisher’s exact test was performed for categorical variables.
The results of the multivariable logistic regression analysis after PS matching are shown in Table 4. The adjusted OR (95% CI) for in-hospital mortality with blood culture test was 0.73 (0.68–0.79). The E-value for in-hospital mortality point estimate was high (2.08), as was the E-value for the 95% CI (1.85). These values indicated that it is unlikely that an unmeasured variable explained the observed association between blood culture test and in-hospital mortality.
| Outcome | Adjusted OR | 95% CI | p-Value | E-Value | Confidence interval |
|---|---|---|---|---|---|
| Primary outcome | |||||
| Death | 0.73 | 0.68–0.79 | <0.0001 | 2.08 | 1.85 |
| Secondary outcome | |||||
| ICU admission | 2.13 | 1.76–2.56 | <0.0001 | 3.68 | 2.92 |
| Ventilator use | 1.14 | 1.04–1.24 | 0.0035 | 1.54 | 1.24 |
To calculate OR, candidate predictors were selected according to a literature review and clinical expertise. We selected 5 covariates (binary variable: blood culture test, sex, age [>65 years], A-DROP score [severe: 4 or 5], and Secondary disease [all shown in Table 2]). OR, odds ratios, CI, confidence intervals.
The results of this study elucidate that performing a blood culture on the first day of hospitalization for CAP inpatients was associated with reduced mortality. To the best of our knowledge, this is the largest epidemiological study conducted to date assessing the usefulness of blood culture in Japanese inpatients with CAP.
The results of the fact-finding survey (Table 1) were not much different from the results of past multicenter joint research in Japan.19) These results suggests that doctors are more likely to request blood cultures in older adults and patients with severe CAP (Table 2).
In this study, we used the data obtained before the publication of the Japanese Respiratory Society (JRS) guidelines for the management of pneumonia in adults in 2017.1) Therefore, it can be inferred that doctors’ decision to perform blood cultures in these patients was not related to the recommendations of the guidelines. The results of this study also revealed that the doctors’ decision to perform blood cultures in these patient groups was warranted.
Waterer et al.20) reported that blood cultures rarely result in an appropriate change in empirical therapy because blood specimens that include skin contaminants can generate false-positive test results. Costantini et al.21) conducted a study of adults hospitalized with CAP and found that blood cultures were associated with a significant increase in the length of hospital stay and duration of antibiotic therapy. The increased use of antibiotics in blood culture groups in our studies may be due to inappropriate antibiotic changes due to false-positive blood culture test results, as shown in these studies.20,21)
One systematic review reported that the proportion of pneumococcal pneumonia which presented as bacteremia ranged from 2.2 to 50.9% (median 28.9%, IQR 14.8 to 33.4%).22)
In Japan, bacteremia was reported in 5.4% of patients hospitalized for CAP23) and 12% of those hospitalized for community-acquired pneumococcal pneumonia.24)
In addition, the main pathogen detected in patients with CAP and bacteremia is Streptococcus pneumoniae.1,23) Ratio of penicillin-susceptible S. pneumoniae was 99.5%25) and if detected early, the chances of successful treatment are high.
These data, as well as the results of our study, support the conclusion that most patients admitted for CAP require blood cultures.
This study has some limitations. First, the NDB does not include claims data from patients with full publicly funded healthcare or for patients who pay for all medical expenses personally. Second, we could not obtain data on the diagnosis or detailed clinical information regarding the pathogens and their drug sensitivity, which prohibited us from evaluating the treatment process, such as the adequacy of antibiotics used. Finally, although by performing PS matching the confounding effect of the covariates could be minimized, the effects of unobserved covariates cannot be adjusted using the PS matching completely. However, the E-value for in-hospital mortality point estimate was high (2.08). Because we have appropriately selected explanatory variables that affect inpatient mortality according to literature reviews and clinical expertise, we believe that, most likely, there are no other covariates that affect both blood culture testing and in-hospital mortality at an OR of 2.08. A nation-wide clinical study needs to be conducted to verify the results of this research.
In conclusion, the results of this study elucidate that performing a blood culture on the first day of hospitalization in CAP inpatients was associated with reduced mortality. The results have great impact in clinical practice. In Japan, a medical database containing clinical laboratory test value information is being developed. Although there are currently a limited number of stored cases, it will be a useful source of information for clinical research in future. We are also considering conducting verification using a large-scale medical information database that includes test value information.
The authors declare no conflict of interest.
