Procedure-based severity index for inpatients: development and validation using administrative database

Identifying inpatient hospitalizations with continuous electroencephalogram monitoring from administrative data

BACKGROUND

Continuous electroencephalography (cEEG) is increasingly utilized in hospitalized patients to detect and treat seizures. Epidemiologic and observational studies using administrative datasets can provide insights into the comparative and cost effectiveness of cEEG utilization. Defining patient cohorts that underwent acute inpatient cEEG from administrative datasets is limited by the lack of validated codes differentiating elective epilepsy monitoring unit (EMU) admissions from acute inpatient hospitalization with cEEG utilization. Our aim was to develop hospital administrative data-based models to identify acute inpatient admissions with cEEG monitoring and distinguish them from EMU admissions.

METHODS

This was a single center retrospective cohort study of adult (≥ 18 years old) inpatient admissions with a cEEG procedure (EMU or acute inpatient) between January 2016-April 2022. The gold standard for acute inpatient cEEG vs. EMU was obtained from the local EEG recording platform. An extreme gradient boosting model was trained to classify admissions as acute inpatient cEEG vs. EMU using administrative data including demographics, diagnostic and procedure codes, and medications.

RESULTS

There were 9,523 patients in our cohort with 10,783 hospital admissions (8.5% EMU, 91.5% acute inpatient cEEG); with average age of 59 (SD 18.2) years; 46.2% were female. The model achieved an area under the receiver operating curve of 0.92 (95% CI [0.91-0.94]) and area under the precision-recall curve of 0.99 [0.98-0.99] for classification of acute inpatient cEEG.

CONCLUSIONS

Our model has the potential to identify cEEG monitoring admissions in larger cohorts and can serve as a tool to enable large-scale, administrative data-based studies of EEG utilization.

Impact of COVID-19 infection rates on admissions for ambulatory care sensitive conditions: nationwide difference-in-difference design in Japan

OBJECTIVES

SARS-CoV-2 infection (COVID-19) has affected tertiary medical institutions and primary care. Admission for ambulatory care sensitive conditions (ACSCs) is an important indicator of primary care quality. However, no nationwide study, especially in Asia, has examined the association between admissions for ACSCs and local surges in COVID-19. This study aimed to examine how the number of admissions for ACSCs has changed in Japan between the areas with higher and lower rates of COVID-19 infection.

DESIGN

This was a retrospective two-stage cross-sectional study. We employed a difference-in-difference design to compare the number of hospital admissions for ACSCs between the areas with higher and lower rates of COVID-19 infection in Japan.

SETTING

The study used a nationwide database in Japan.

PARTICIPANTS

All patients were aged 20 years and above and were admitted due to ACSCs during the study period between March and September 2019 (before the pandemic) and between March and September 2020 (during the pandemic).

RESULTS

The total number of ACSC admissions was 464 560 (276 530 in 2019 and 188 030 in 2020). The change in the number of admissions for ACSCs per 100 000 was not statistically significant between the areas with higher and lower rates of COVID-19 infection: 7.50 (95% CI -87.02 to 102.01). In addition, in acute, chronic and preventable ACSCs, the number of admissions per 100 000 individuals did not change significantly.

CONCLUSION

Although admissions for ACSCs decreased during the COVID-19 pandemic, there was no significant change between the areas with higher and lower rates of COVID-19 infection. This implies that the COVID-19 pandemic affected the areas with higher infection rates and the areas with lower rates.

Gabapentinoid Use Is Associated With Reduced Occurrence of Hyperactive Delirium in Older Cancer Patients Undergoing Chemotherapy: A Nationwide Retrospective Cohort Study in Japan

BACKGROUND

It is unclear whether gabapentinoids affect the development of delirium. We aimed to determine the association between gabapentinoid use and hyperactive delirium in older cancer patients undergoing chemotherapy.

METHODS

We conducted propensity score-matched analyses using data from a nationwide inpatient database in Japan. We included cancer patients with pain ≥70 years of age undergoing chemotherapy between April 2016 and March 2018. Patients receiving gabapentinoids were matched with control patients using propensity scores. The primary outcome was occurrence of hyperactive delirium during hospitalization, and the secondary outcomes were length of hospital stay, in-hospital fractures, and in-hospital mortality. Hyperactive delirium was identified by antipsychotic use or discharge diagnoses from the International Classification of Diseases, 10th Revision.

RESULTS

Among 143,132 identified patients (59% men; mean age, 76.3 years), 14,174 (9.9%) received gabapentinoids and 128,958 (90.1%) did not (control group). After one-to-one propensity score matching, 14,173 patients were included in each group. The occurrence of hyperactive delirium was significantly lower (5.2% vs 8.5%; difference in percent, -3.2% [95% confidence interval, -3.8 to -2.6]; odds ratio, 0.60 [0.54-0.66]; P < .001), the median length of hospital stay was significantly shorter (6 days [interquartile range, 3-15] vs 9 days [4-17]; subdistribution hazard ratio, 1.22 [1.19-1.25]; P < .001), and the occurrence of in-hospital mortality was significantly lower in the gabapentinoid group than in the control group (1.3% vs 1.8%; difference in percent, -0.6% [-0.9 to -0.3]; odds ratio, 0.69 [0.57-0.83]; P < .001). Gabapentinoid use was not significantly associated with the occurrence of in-hospital fractures (0.2% vs 0.2%; difference in percent, 0.0% [-0.1 to 0.1]; odds ratio, 1.07 [0.65-1.76]; P = .799). The results of sensitivity analyses using stabilized inverse probability of treatment weighting were consistent with the results of the propensity score-matched analyses.

CONCLUSIONS

Our findings suggest that gabapentinoid use is associated with reduced hyperactive delirium in older cancer patients undergoing chemotherapy, with no evidence of an increase in the fracture rate, length of hospital stay, or in-hospital death.

Development of Deep Learning Models for Predicting In-Hospital Mortality Using an Administrative Claims Database: Retrospective Cohort Study

BACKGROUND

Administrative claims databases have been used widely in studies because they have large sample sizes and are easily available. However, studies using administrative databases lack information on disease severity, so a risk adjustment method needs to be developed.

OBJECTIVE

We aimed to develop and validate deep learning-based prediction models for in-hospital mortality of acute care patients.

METHODS

The main model was developed using only administrative claims data (age, sex, diagnoses, and procedures on the day of admission). We also constructed disease-specific models for acute myocardial infarction, heart failure, stroke, and pneumonia using common severity indices for these diseases. Using the Japanese Diagnosis Procedure Combination data from July 2010 to March 2017, we identified 46,665,933 inpatients and divided them into derivation and validation cohorts in a ratio of 95:5. The main model was developed using a 9-layer deep neural network with 4 hidden dense layers that had 1000 nodes and were fully connected to adjacent layers. We evaluated model discrimination ability by an area under the receiver operating characteristic curve (AUC) and calibration ability by calibration plot.

RESULTS

Among the eligible patients, 2,005,035 (4.3%) died. Discrimination and calibration of the models were satisfactory. The AUC of the main model in the validation cohort was 0.954 (95% CI 0.954-0.955). The main model had higher discrimination ability than the disease-specific models.

CONCLUSIONS

Our deep learning-based model using diagnoses and procedures produced valid predictions of in-hospital mortality.

Use of naldemedine is associated with reduced incidence of hyperactive delirium in cancer patients with opioid-induced constipation: a nationwide retrospective cohort study in Japan

BACKGROUND

Medical benefits of peripherally-acting mu-opioid receptor antagonists other than improving opioid-induced constipation remain unclear. Our aim was to evaluate the association between the use of naldemedine and incidence of hyperactive delirium in cancer patients receiving chemotherapy and opioid therapy.

METHODS

We conducted a propensity score-matched analysis using a nationwide inpatient database in Japan. Cancer patients receiving both inpatient chemotherapy and opioid therapy from June 1, 2017 to March 31, 2018 were included. Patients receiving naldemedine were matched to control patients by propensity score. Our primary outcome was the incidence of hyperactive delirium during hospitalization, and secondary outcomes were the length of hospital stay, hospital costs, in-hospital mortality, and incidence of ileus.

RESULTS

Of 34,031 patients receiving inpatient chemotherapy and opioid therapy, 1905 (5.6%) were included in the naldemedine group. After one-to-four propensity score matching, 1904 patients were included in the naldemedine group and 7616 in the control group. Naldemedine users had significantly reduced incidence of hyperactive delirium compared with the control patients (19.4% vs 23.3%; risk difference, -3.9 [95% confidence interval, -5.9 - -1.9]; risk ratio, 0.83 [0.75-0.92]; P<0.001; subdistribution hazard ratio, 0.85 [0.75-0.97]; P=0.015). The median length of hospital stay was significantly shorter in the naldemedine group compared with the control group (12 days [interquartile range, 6-23] vs 14 days [6-26]; P=0.001). The median hospital costs were also significantly lower in the naldemedine group compared with the control group (US $6179 [3351-10,026] vs US $6576 [3436-11,107]; P<0.001). No significant differences were found for in-hospital mortality or incidence of ileus between the groups.

CONCLUSIONS

Our findings suggest that the use of naldemedine may have benefits in preventing hyperactive delirium, shortening hospital stay, and decreasing hospital costs in cancer patients receiving chemotherapy and opioid therapy.

Efficacy of Prophylactic Antibiotics during Extracorporeal Membrane Oxygenation: A Nationwide Cohort Study

RATIONALE

Extracorporeal membrane oxygenation (ECMO) is widely used worldwide, and many hospitals consider using antibiotics to prevent nosocomial infection in ECMO patients. However, the efficacy of antimicrobial prophylaxis for patients receiving ECMO remains unclear.

OBJECTIVE

This study aimed to evaluate the efficacy of prophylactic antibiotics for ECMO patients.

METHODS

This retrospective cohort study was performed using data recorded in the Japanese Diagnosis Procedure Combination database from July 1, 2010, to March 31, 2017. We compared in-hospital mortality, nosocomial pneumonia, acute kidney injury, and diarrhea between patients receiving prophylactic antibiotics (prophylaxis group) and those not receiving antibiotics (control group) using propensity score matching, multivariable logistic regression models fitted using generalized estimating equation, stabilized inverse probability of treatment weighting, and instrumental variable analysis.

RESULTS

We identified 9615 eligible patients and classified them into the prophylaxis (n = 5552) and control (n = 4063) groups. In the propensity score-matched analysis, 3650 pairs were generated. Significant differences between the prophylaxis group and the control group were detected in in-hospital mortality (56.4% vs 59.8%, risk difference: -3.7%, 95% confidence interval [CI]: -6.0 to -1.3) and nosocomial pneumonia (12.9% vs. 15.3%, risk difference: -2.4%, 95% CI: -4.3 to -0.6). In the multivariable logistic regression models fitted using generalized estimating equation, stabilized inverse probability of treatment weighting, and instrumental variable analysis, the point estimates a suggested similar direction to that found in the propensity score analysis. The proportions of patients with acute kidney injury or diarrhea did not significantly increase with the use of prophylactic antibiotics.

CONCLUSION

Using prophylactic antibiotics during ECMO was associated with reduced in-hospital mortality and lower proportions of patients with nosocomial pneumonia. Future prospective studies are needed to validate these results.

Development and validation of a procedure-based organ failure assessment model for patients in the intensive care unit: an administrative database study

Aim

To develop a procedure‐based organ failure assessment model for intensive care unit (ICU) patients and to examine the ability of this model to predict in‐hospital mortality, with reference to the Sequential Organ Failure Assessment (SOFA) score.

Methods

Using the Japanese nationwide Diagnosis Procedure Combination database, we identified patients aged ≥15 years who were admitted to the ICUs April 2018–March 2019. Since April 2018, Japanese health care providers have been required to input ICU patients' SOFA scores into this database. We extracted data on the following procedures on ICU admission: oxygen supplementation, invasive mechanical ventilation, blood transfusions, catecholamines, chest compression, extracorporeal membrane oxygenation, and renal replacement therapy. A procedure‐based organ failure assessment model (Model 1) for in‐hospital mortality was developed using therapeutic procedures for organ failure on the day of ICU admission in the derivation cohort. We also constructed a model using the SOFA score (Model 2). Discriminatory ability was assessed using area under the receiver operating characteristic curve (AUROC) in the validation cohort, and the discriminatory abilities of the models were compared.

Results

In total, 69,019 patients were included. Overall in‐hospital mortality was 7.2%. The AUROCs for Model 1 (0.810) and Model 2 (0.817) in the validation cohort did not show a statistically significant difference (P = 0.20).

Conclusion

The models established using procedure‐based organ failure assessment showed no statistically significant differences from those using the SOFA score, suggesting that procedure records in administrative databases can be used for risk adjustment in clinical studies on ICU mortality.

Proton pump inhibitors versus histamine-2 receptor antagonists for stress ulcer prophylaxis during extracorporeal membrane oxygenation: a propensity score-matched analysis

PURPOSE

Patients receiving extracorporeal membrane oxygenation (ECMO) generally receive proton pump inhibitors (PPIs) or histamine-2 receptor antagonists (H₂RAs) to avoid major gastrointestinal bleeding. Our aim was to compare outcomes between patients receiving PPIs and H₂RAs for stress ulcer prophylaxis during ECMO.

MATERIALS AND METHODS

We performed a retrospective cohort study using the Japanese Diagnosis Procedure Combination Database, using data recorded from 1 July 2010 to 31 March 2017. We defined patients who received PPIs within 2 days after starting ECMO as the PPIs group and those who received H₂RAs within 2 days after starting ECMO as the H₂RAs group. We performed propensity score matching to compare outcomes. The primary outcomes were gastrointestinal bleeding requiring endoscopic haemostasis and in-hospital mortality. The secondary outcomes were red blood cell transfusion, hospital-acquired pneumonia and Clostridium difficile infection during hospitalisation.

RESULTS

Of 11 328 eligible patients, 9738 received PPIs and 1590 received H₂RAs. Propensity score matching created 1556 pairs. No significant differences were seen regarding endoscopic haemostasis (1.2% vs 0.8%; p=0.37), in-hospital mortality (53.0% vs 53.1%; p=0.94), red blood cell transfusion rates (91.4% vs 89.7%; p=0.11), hospital-acquired pneumonia (13.0% vs 12.4%; p=0.59) or C. difficile infection (0.1% vs 0.2%; p=0.32) between the PPIs and H₂RAs groups, respectively.

CONCLUSION

We found no differences in the evaluated outcomes between the PPIs and H₂RAs groups. Both PPIs and H₂RAs are treatment options for stress ulcer prophylaxis in patients undergoing ECMO.

Comparing outcomes after peripheral nerve block versus general anesthesia for lower extremity amputation: a nationwide exploratory retrospective cohort study in Japan

Background and objectives

The health benefits of peripheral nerve block (PNB) on postoperative complications after lower extremity amputation (LEA) compared with general anesthesia (GA) remains controversial. We performed a retrospective propensity score-matched cohort analysis to compare major outcomes after LEA with PNB versus GA.

Materials and methods

We used a nationwide inpatient database in Japan to compare patient outcomes after LEA with PNB versus GA from 2010 to 2016. Our primary outcome was 30-day mortality after LEA. The incidence of composite morbidity from life-threatening complications and of delirium within 30 days after LEA were secondary outcomes. We conducted propensity score-matched analyses of patients who underwent below knee or foot amputation using 36 covariates. Logistic regression analyses fitted with generalized estimating equations were performed to calculate ORs and their 95% CIs.

Results

Of 11 796 patients, 747 received PNB and 11 049 received GA. After one-to-four propensity score matching, 747 patients were included in the PNB group and 2988 in the GA group. The adjusted ORs for postoperative mortality, composite morbidity and delirium within 30 days after LEA were 1.11 (95% CI 0.75 to 1.64), 1.15 (95% CI 0.85 t o1.56) and 0.75 (95% CI 0.57 to 0.98), respectively, for the PNB group with reference to the GA group.

Conclusions

There was no significant difference between groups in 30-day mortality or composite morbidity. The PNB group showed a significantly lower risk of postoperative delirium than the GA group. Our findings suggest that PNB may have advantages over GA in preventing postoperative delirium among patients undergoing LEA.

Nationwide observational study of mortality from complicated intra-abdominal infections and the role of bacterial cultures

BACKGROUND

The benefit of taking intra-abdominal cultures during source control procedures in patients with complicated intra-abdominal infection (CIAI) is unknown. The aim of this study was to evaluate whether intra-abdominal cultures reduce the mortality rate of CIAI.

METHODS

The Japanese Diagnosis Procedure Combination database was used to identify adult patients with CIAI who had undergone source control procedures on the first day of admission to hospital between April 2014 and March 2016. In-hospital mortality was compared between patients who did and those who did not have intra-abdominal cultures taken. A generalized linear mixed-effect logistic regression model and a random intercept per hospital were used to adjust for baseline confounders and institutional differences. Subgroup analyses were also performed according to disease cause, site of onset and severity of CIAI.

RESULTS

Intra-abdominal cultures were taken from 16 303 of 41 495 included patients. Multivariable logistic regression analysis showed that patients with intra-abdominal cultures had a significantly lower mortality than those without (odds ratio 0·85, 95 per cent c.i. 0·77 to 0·95). Subgroup analyses revealed statistically significant differences in mortality between patients with and without cultures among those with lower intestinal perforation, biliary tract infection/perforation, healthcare-associated CIAI and high-risk community-acquired CIAI.

CONCLUSIONS

Intra-abdominal cultures obtained during source control procedures may reduce in-hospital mortality, especially in patients with lower intestinal perforation, biliary tract infection/perforation, or healthcare-associated or high-risk community-acquired CIAI.

Incidence and Age Distribution of Hospitalized Presumptive and Possible Abusive Head Trauma of Children Under 12 Months Old in Japan

Background

Abusive head trauma (AHT) is the leading cause of fatal maltreatment among young children. The incidence of AHT in Japan, however, remains unknown. This study examined the incidence and distribution of age in months among young children under 12 months of age hospitalized with intracranial injury in Japan.

Methods

We conducted a multicenter cross-sectional study of children under 36 months old admitted with intracranial injury to hospitals that employed the Diagnostic Procedure Combination (DPC) payment system between 2010 and 2013. Presumptive and possible AHT were defined using the combination of ICD-10 codes modified from the coding system recommended by the United States Centers for Disease Control and Prevention.

Results

The average incidence was 7.2 (95% confidence interval [CI], 7.18–7.26) for presumptive and 41.7 (95% CI, 41.7–41.8) for possible AHT per 100,000 children less than 12 months old from 2010 to 2013. The distributions of age in months for both presumptive AHT and possible AHT had peaks at around 2 and 8 months.

Conclusions

This is the first study to report the incidence of hospitalized children with presumptive and possible AHT using population-based data. Further datasets are needed to evaluate the incidence and specific preventive strategies to prevent AHT in infants during the months of highest risk.

Scalable and accurate deep learning with electronic health records

Predictive modeling with electronic health record (EHR) data is anticipated to drive personalized medicine and improve healthcare quality. Constructing predictive statistical models typically requires extraction of curated predictor variables from normalized EHR data, a labor-intensive process that discards the vast majority of information in each patient's record. We propose a representation of patients' entire raw EHR records based on the Fast Healthcare Interoperability Resources (FHIR) format. We demonstrate that deep learning methods using this representation are capable of accurately predicting multiple medical events from multiple centers without site-specific data harmonization. We validated our approach using de-identified EHR data from two US academic medical centers with 216,221 adult patients hospitalized for at least 24 h. In the sequential format we propose, this volume of EHR data unrolled into a total of 46,864,534,945 data points, including clinical notes. Deep learning models achieved high accuracy for tasks such as predicting: in-hospital mortality (area under the receiver operator curve [AUROC] across sites 0.93-0.94), 30-day unplanned readmission (AUROC 0.75-0.76), prolonged length of stay (AUROC 0.85-0.86), and all of a patient's final discharge diagnoses (frequency-weighted AUROC 0.90). These models outperformed traditional, clinically-used predictive models in all cases. We believe that this approach can be used to create accurate and scalable predictions for a variety of clinical scenarios. In a case study of a particular prediction, we demonstrate that neural networks can be used to identify relevant information from the patient's chart.

Association between mode of anaesthesia and severe maternal morbidity during admission for scheduled Caesarean delivery: a nationwide population-based study in Japan, 2010-2013

BACKGROUND

Although the incidence of maternal mortality during Caesarean delivery remains very low, the rate of severe maternal morbidity is increasing. Improvements in obstetric anaesthetic practice have resulted in a dramatic reduction in the risk of maternal death from general anaesthesia. Less clear is whether the risk of severe maternal morbidity differs according to mode of anaesthesia for women undergoing Caesarean delivery. We analysed the association between the mode of anaesthesia and severe maternal morbidity during Caesarean delivery using a nationally representative inpatient database.

METHODS

We identified 89 225 women undergoing scheduled Caesarean delivery from the Diagnosis Procedure Combination database in Japan, 2010-2013. We defined severe maternal morbidity as the presence of any life-threatening complications and identified women with severe maternal morbidity from the database. Propensity score-matched analysis was carried out to compare the odds of severe maternal morbidity between women who underwent general vs neuraxial anaesthesia.

RESULTS

Of 89 225 women, 10 058 received general anaesthesia and 79 167 received neuraxial anaesthesia. In the propensity score-matched analysis with 10 046 pairs, a higher incidence of severe maternal morbidity was observed among patients receiving general (2.00%) rather than neuraxial anaesthesia (0.76%). The odds ratio of severe maternal morbidity was 2.68 (95% CI, 1.97-3.64) among women receiving general compared with neuraxial anaesthesia.

CONCLUSIONS

For scheduled Caesarean delivery, general anaesthesia compared with neuraxial anaesthesia is associated with greater odds for severe maternal morbidity. However, we should be cautious with interpretation of these findings because they may be explained by confounding indications.

Validity of diagnoses, procedures, and laboratory data in Japanese administrative data

Background

Validation of recorded data is a prerequisite for studies that utilize administrative databases. The present study evaluated the validity of diagnoses and procedure records in the Japanese Diagnosis Procedure Combination (DPC) data, along with laboratory test results in the newly-introduced Standardized Structured Medical Record Information Exchange (SS-MIX) data.

Methods

Between November 2015 and February 2016, we conducted chart reviews of 315 patients hospitalized between April 2014 and March 2015 in four middle-sized acute-care hospitals in Shizuoka, Kochi, Fukuoka, and Saga Prefectures and used them as reference standards. The sensitivity and specificity of DPC data in identifying 16 diseases and 10 common procedures were identified. The accuracy of SS-MIX data for 13 laboratory test results was also examined.

Results

The specificity of diagnoses in the DPC data exceeded 96%, while the sensitivity was below 50% for seven diseases and variable across diseases. When limited to primary diagnoses, the sensitivity and specificity were 78.9% and 93.2%, respectively. The sensitivity of procedure records exceeded 90% for six procedures, and the specificity exceeded 90% for nine procedures. Agreement between the SS-MIX data and the chart reviews was above 95% for all 13 items.

Conclusion

The validity of diagnoses and procedure records in the DPC data and laboratory results in the SS-MIX data was high in general, supporting their use in future studies.

•

Validity of DPC and SS-MIX data was tested using chart review results as references.

•

Specificity of DPC diagnoses was high; sensitivity was low and varied by conditions.

•

Procedure records in DPC data were highly sensitive and specific.

•

Accuracy of SS-MIX laboratory data exceeded 95% in all 13 items examined.

Increase in avoidable hospital admissions after the Great East Japan Earthquake

BACKGROUND

The Great East Japan Earthquake and subsequent tsunami and nuclear disaster on 11 March 2011 had a short-term influence on the increase in emergency department visits and hospital admissions due to various diseases. However, it remains unclear whether the earthquake and tsunami disaster affected the long-term health conditions of people in the affected areas.

METHODS

Using a national inpatient database in Japan, we investigated people's ambulatory care sensitive conditions (ACSCs), which are defined as conditions for which effective management and treatment should prevent admission to a hospital. We compared the number of admissions for ACSCs before-quake (July 2010 to February 2011) with after-quake (July 2012 to February 2013) periods in the disaster area compared with other areas using a difference-in-differences design. Linear regression models with the interaction between periods and areas were used to estimate the impact of the earthquake on admissions for ACSCs.

RESULTS

No significant difference in difference was seen in preventable ACSCs (where immunisation and other interventions can prevent illness) or chronic ACSCs (where effective care can prevent flare-ups), while acute ACSCs (where early intervention can prevent more serious progression) increased significantly (3.3 admissions per 100 000 population; 95% CI 0.4 to 6.3; p=0.028).

CONCLUSIONS

Preventable and chronic ACSCs may have increased just after the earthquake and then immediately decreased. However, avoidable admissions due to acute ACSCs remained high in the long term after the earthquake and tsunami disaster.

De-escalation versus continuation of empirical antimicrobial therapy in community-acquired pneumonia

OBJECTIVES

To compare mortality between de-escalation and continued empirical treatment in patients with community-acquired pneumonia.

METHODS

Using a nationwide administrative database, we identified adult patients with community-acquired pneumonia caused by Streptococcus pneumoniae, other streptococci, Haemophilus influenzae, Klebsiella pneumoniae, or Escherichia coli (n = 10,231) or of unknown etiology (n = 8247), discharged between July 2010 and March 2013. De-escalation was determined by the spectrum and number of antimicrobials at day 4. We used propensity score matching to obtain 489 pairs of de-escalation and continuation groups among pathogen-identified patients and 278 pairs among culture-negative patients to compare mortalities.

RESULTS

In the pathogen-identified patients, de-escalation was noninferior to continuation in 15-day mortality [5.3% in de-escalation versus 4.3% in continuation, a difference of 1.0% (95% confidence interval, -1.7% to 3.7%)] and in-hospital mortality [8.0% in de-escalation versus 8.8% in continuation, a difference of -0.8% (95% confidence interval, -4.3% to 2.7%)]. In the culture-negative cases, de-escalation was noninferior to continuation in terms of 15-day mortality but not in terms of in-hospital mortality.

CONCLUSIONS

Among patients with community-acquired pneumonia of specific etiology, de-escalation was noninferior to continuation of empirical treatment, suggesting that de-escalation is a safe strategy and supporting current recommendations. Safety of de-escalation in culture-negative cases is questionable.