Lower respiratory tract infection (LRTI) burden estimates increasingly rely on administrative databases using International Classification of Diseases (ICD) codes, but no standard methodology exists. |
Our systematic review of validation studies showed that most ICD code-based algorithms miss a substantial number of hospitalized LRTI cases in adults, whether community- or hospital-acquired cases. |
ICD-based algorithms containing only nosocomial- or pathogen-specific ICD codes to identify hospital-acquired and pneumococcal pneumonia have poor sensitivity. |
ICD-based algorithms should include all pneumonia ICD codes, in particular those for unspecified organisms, to estimate the incidence of all-cause hospitalized LRTI and hospital-acquired pneumonia in adults. |
Introduction
Methods
Study Selection and Data Extraction
Data Analysis
Best Practices
Results
Study Description
Country (study) | Patient population | Algorithm and coding position* | Pneumonia codes | Aspiration pneumonia | Other infection codes (that are not pneumonia codes) | Other codes of respiratory system | Reference standard | Se (%) | Sp (%) | LR+ | LR− |
---|---|---|---|---|---|---|---|---|---|---|---|
Any LRTI | |||||||||||
ICD-10 | |||||||||||
Denmark (Henriksen, 2014) | ≥ 15 years | ICD-10, any position | J12–18, without influenza and some missing | Yes (J69) | Yes, bacterial and fungal pneumonia, opportunistic LRTI, tuberculosis | Yes, bronchitis, lung abscess, COPD with LRTI, pyothorax | Chart review (clinical, radiological and lab) | 71 | 92 | 8 | 0.32 |
Thailand (Rattanaumpawan, 2016) | ≥ 15 years | ICD-10, any position | J12–18, without influenza | No | Yes, bacterial, viral, fungal, parasitic pneumonia and tuberculosis | No | Chart review (clinical and lab) | 57 | 94 | 9 | 0.46 |
Pneumonia overall | |||||||||||
ICD-9 | |||||||||||
USA (Aronsky, 2005) | ≥ 18 years | ICD-9 codes in primary | 480–487, without infectious diseases classified elsewhere (484) | No | No | No | Chart review (clinical and radiological) and recoding | 55 | 99 | 61 | 0.46 |
ICD-9 codes in primary | 480–487 | Yes (507) | Yes, bacterial, viral, fungal, toxoplasmosis pneumonia | Yes, empyema and other unspecified pneumonia-related | 68 | 99 | 68 | 0.32 | |||
ICD-9 primary, or secondary with codes of pneumonia severity as primary | 480–487, without infectious diseases classified elsewhere (484) | Yes (507) | No** | No** | 70 | 99 | 63 | 0.31 | |||
USA (Ahmed, 2014) | ≥ 18 years, patients at risk of ARDS | ICD-9 codes, position not stated | 480–486, without influenza | No | No | No | Chart review and concordance between three algorithms | 77 | 98 | 39 | 0.23 |
ICD-10 | |||||||||||
Australia (Skull, 2008) | ≥ 65 years | ICD-10, any position | J10–18 | No | No | No | Radiological data | 89 | 62 | 2 | 0.17 |
Chart review | 98 | 97 | 32 | 0.02 | |||||||
Chart review and radiological data | 98 | 64 | 3 | 0.03 | |||||||
Community-acquired pneumonia | |||||||||||
ICD-9 | |||||||||||
USA (Whittle, 1997) | ≥ 18 years, excluding trauma and elective surgery patients | ICD-9, primary | 480–487 | Yes (507) | Yes, bacterial, viral, fungal, toxoplasmosis pneumonia | Yes, empyema and other unspecified pneumonia-related | Chart review (clinical and radiological) | 84 | 86 | 6 | 0.19 |
ICD-9 primary, or secondary with codes of pneumonia severity as primary, or “pneumonia” in patient classification | 480–487 | 89 | 80 | 4 | 0.14 | ||||||
Singapore (Mukhopadhyay, 2017) | ≥ 21 years | ICD-9, position not stated + text mining for pneumonia in admission and in X-ray report | 480–487 | No | No | No | Diagnosis of reviewing physician based on chart review | 85 | 78 | 4 | 0.20 |
Netherlands (van de Garde, 2007) | ≥ 18 years | ICD-9, primary | 481–486, without viral pneumonia and influenza | No | No | No | Chart review (clinical, laboratory and radiological) | 72 | |||
ICD-9, any position | 481–486, without viral pneumonia and influenza | No | No | No | Chart review (clinical, laboratory and radiological) | 80 | |||||
USA (Yu, 2011) | 18–64 years | ICD-9, primary | 480–487 | Yes (507) | No | No | Diagnosis by treating physician | 63 | 93 | 9 | 0.40 |
≥ 65 years | ICD-9, primary | No | 65 | 85 | 4 | 0.41 | |||||
18–64 years | ICD-9 pneumonia codes primary OR LOS ≤ 3 days without 507 OR LOS > 3 days with ARF in primary | 480–487 | Excluded (507) | No | No** | Diagnosis by treating physician | 81 | 82 | 5 | 0.23 | |
≥ 65 years | ICD-9 primary OR LOS ≤ 3 days without 507 OR LOS > 3 days with ARF in primary OR no pneumonia codes in primary nor 507.0 in any field, with LOS ≤ 9 days without an operation procedure code | 480–487 | Excluded (507) | No | No** | Diagnosis by treating physician | 89 | 63 | 2 | 0.17 | |
USA (Jones, 2018) | ≥ 18 years seen at ED | ICD-9 primary, or secondary with RF or sepsis in primary, ED | 481–486, without viral pneumonia and influenza | No | No** | No** | Diagnosis or treatment by treating physician | 44 | 100 | ∞ | 0.56 |
≥ 18 years hospitalized | ICD-9 primary, or secondary with RF or sepsis in primary, discharge | No** | No** | Diagnosis or treatment by treating physician | 67 | 99 | 67 | 0.33 | |||
≥ 18 years seen at ED | ICD-9 primary, or secondary with sepsis or RF in primary in ED + natural language processing | 481–486, without viral pneumonia and influenza | No | No** | No** | Diagnosis or treatment by treating physician | 89 | 98 | 45 | 0.11 |
Characteristics and Accuracy of ICD Algorithms
Any or Community-Acquired LRTI, All Causes
Hospital-Acquired Pneumonia, All Causes
Country (study) | Patient population | Algorithm and coding position* | Pneumonia codes | Aspiration pneumonia | Other infection codes (that are not pneumonia codes) | Other codes of respiratory system | Reference standard | Se (%) | Sp (%) | LR+ | LR− |
---|---|---|---|---|---|---|---|---|---|---|---|
Any hospital-acquired pneumonia | |||||||||||
ICD-9 | |||||||||||
Canada (Quan, 2004) | ≥ 18 years | ICD-9, any position (up to 16) | 480–486, without influenza | Yes (507) | Yes, bacterial, viral, fungal, parasitic pneumonia and tuberculosis | Yes, empyema and other unspecified pneumonia-related | Chart review using recoding | 35 | 99 | 58 | 0.65 |
USA (Romano, 2002) | ≥ 18 years with elective lumbar diskectomy | ICD-9, position not specified | 480–487 | Yes (507) | No | Yes, empyema and other unspecified pneumonia-related | Chart review using recoding | 100 | 100 | 333 | 0.00 |
ICD-10 | |||||||||||
Germany (Azaouagh, 2008) | ≥ 18 years with mixed comorbidities | ICD-10 codes not present on admission | J10–18 | Viral and bacterial pneumonia specific codes | Other affections related to respiratory system | Chart review (clinical, radiological and laboratory data) | 43 | 99 | 44 | 0.58 | |
Switzerland (Wolfensberger, 2018) | ≥ 18 years | ICD-10 code for HAP | Code for HAP (U69.00) | No | No | No | Chart review (clinical, radiology and lab—HELICS definition) | 59 | 98 | 33 | 0.42 |
Germany (Maass, 2015) | ≥ 65 years | ICD-10 codes not present on admission and nosocomial origin, any position | Code for HAP (U69.00. German version) | No | No | No | Chart review, according to CDC criteria | 23 | 100 | 233 | 0.77 |
Ventilator-associated pneumonia | |||||||||||
ICD-9 | |||||||||||
USA (Cass, 2013) | ≥ 18 years | ICD-9 for ventilator-associated pneumonia, primary, not present on admission (restricted) | None | No | No | Ventilator-associated pneumonia | As defined by hospital surveillance | 25 | 100 | 83 | 0.75 |
≥ 18 years | ICD-9 30 diagnosis codes + ventilator-associated pneumonia, not present on admission (expanded) | 480–487, without infectious diseases classified elsewhere (484) | No | No | Ventilator-associated pneumonia | As defined by hospital surveillance | 61 | 93 | 9 | 0.42 | |
≥ 18 years with continuous invasive mechanical ventilation | ICD-9 for ventilator-associated pneumonia, not present on admission (restricted) | None | No | No | Ventilator-associated pneumonia | As defined by hospital surveillance | 25 | 99 | 21 | 0.76 | |
≥ 18 years with continuous invasive mechanical ventilation | ICD-9 30 diagnosis codes + ventilator-associated pneumonia, not present on admission (expanded) | 480–487, without infectious diseases classified elsewhere (484) | No | No | Ventilator-associated pneumonia | As defined by hospital surveillance | 61 | 83 | 4 | 0.47 |
Pathogen-Specific LRTI
Country (study) | Population | Algorithm and coding position* | Pneumonia codes | Aspiration pneumonia | Other infection codes (that are not pneumonia codes) | Other codes of respiratory system | Reference standard | Pathogen | Se (%) | Sp (%) | LR+ | LR− |
---|---|---|---|---|---|---|---|---|---|---|---|---|
Pneumonia | ||||||||||||
USA (Higgins, 2020) | ≥ 18 years | ICD-9 codes in primary, OR secondary with ARF or sepsis in primary, discharge | 482.41 | No | Yes, sepsis and bacteremia codes | Yes, other unspecified pneumonia-related, ARF | Laboratory data (blood or respiratory culture, urinary antigen, or PCR) | Methicillin-sensitive S. aureus | 14 | 100 | 142 | 0.9 |
482.42 | 49 | 99 | 82 | 0.5 | ||||||||
481, 482.30 | S. pneumoniae | 60 | 99 | 75 | 0.4 | |||||||
482.1 | Pseudomonas | 47 | 100 | 156 | 0.5 | |||||||
482.82 | E. coli | 17 | 100 | 0.8 | ||||||||
482 | Klebsiella pneumoniae | 36 | 100 | 357 | 0.6 | |||||||
482.2 | H. influenzae | 43 | 100 | 428 | 0.6 | |||||||
483 | Mycoplasma pneumoniae | 54 | 100 | 268 | 0.5 | |||||||
482.84 | Legionella | 78 | 100 | 784 | 0.2 | |||||||
487.x, 488.x | Influenza | 96 | 99 | 87 | 0.0 | |||||||
480.1 | RSV | 24 | 100 | 0.8 | ||||||||
480.2 | Parainfluenza | 14 | 100 | 0.9 | ||||||||
Community-acquired pneumonia | ||||||||||||
USA (Guevara, 1999) | ≥ 18 years | ICD-9-CM code in primary | 481 | No | No | No | Clinical data, radiological and laboratory data (blood or pleural fluid) | S. pneumoniae | 45 | 98 | 21 | 0.6 |
ICD-9-CM code in 1–5 position | No | No | No | 58 | 98 | 23 | 0.4 | |||||
ICD-9-CM code in primary | 482.3 | No | No | No | 7 | 99 | 8 | 0.9 | ||||
ICD-9-CM code in 1–5 position | No | No | No | 11 | 99 | 11 | 0.9 | |||||
ICD-9-CM code in first position | 486 | No | No | No | 11 | 55 | 0 | 1.6 | ||||
ICD-9-CM code in 1–5 position | No | No | No | 14 | 48 | 0 | 1.8 | |||||
ICD-9-CM codes in 1–5 position | 481.00 | No | Yes, pneumococcal septicemia | No | 64 | 97 | 25 | 0.4 | ||||
481.00 | No | Yes, pneumococcal and streptococcal septicemia | No | 77 | 97 | 25 | 0.2 | |||||
481.00, 482.30 | No | No | 81 | 96 | 21 | 0.2 | ||||||
481.00, 482.30 | No | Yes, ARF* | 81 | 96 | 20 | 0.2 | ||||||
481.00, 482.30, 486.00 | No | Yes, ARF* | 89 | 45 | 2 | 0.2 | ||||||
Netherlands (Van de Garde, 2007) | ≥ 18 years | ICD-9 codes in primary or secondary, admission and discharge | 481 | No | No | No | Clinical, radiological and laboratory data (From sputum and blood, urinary antigen) | S. pneumoniae | 35 | |||
482.x, 483.x | No | No | No | Pneumonia with other organism specified | 18 | |||||||
485–486 | No | No | No | Pneumonia, organism unspecified | 63 |
Best Practices Learned from Included Studies
Distinguishing Community-Acquired from Hospital-Acquired Pneumonia
Algorithms Including Code Position
Discussion
Best practices | Evidence |
---|---|
Do not use specific ICD codes to determine incidence rates of: • Pneumococcal LRTI • Hospital-acquired pneumonia | |
Include pneumonia codes ICD-9 480–488 or ICD-10 J10-18 in primary or any position: • Patients with severe pneumonia and underlying diseases will be better captured when pneumonia codes are in any position • Always include the “organism unspecified” pneumonia code (ICD-9 486 or ICD-10 J18.9) • Prefer to include all pneumonia and influenza codes | |
Sensitivity improves compared to pneumonia codes in primary position, with limited changes to specificity when: | |
• Codes for severe pneumonia (e.g. sepsis, acute respiratory failure, pneumothorax) are added in primary position, with pneumonia codes in secondary position | |
• Other infection codes (e.g. pathogens causing pneumonia) and/or respiratory codes (e.g. empyema, lung abscess) are added | 1 study [11] for both infection or respiratory codes |
• Pneumonia and influenza codes are in any position | 1 study [32], but no data on specificity |
• Free text search for pneumonia terms are added, when feasible | 1 study [19] |
Unclear whether sensitivity would improve or not: • Code for aspiration pneumonia is added or excluded |