The pneumonia severity index (PSI) is a clinical prediction tool that sorts patients with community-acquired pneumonia into five risk classes based on their likelihood of dying within 30 days. It was derived and validated using data from more than 50,000 patients and relies on 20 demographic and clinical variables that are routinely available when someone shows up to an emergency department.1Clinical Infectious Diseases. The Pneumonia Severity Index: A Decade after the Initial Derivation and Validation Its core purpose is not to diagnose pneumonia but to answer a narrower, high-stakes question: can this patient safely go home with oral antibiotics, or do they need a hospital bed? That deceptively simple question has shaped emergency department workflows, guideline recommendations, and health-care costs for decades, though the score’s real-world use is more complicated than the math suggests.
What the Score Actually Measures
The PSI collects information across three broad categories: demographics (age, sex, nursing-home residence), coexisting illnesses (cancer, liver disease, heart failure, kidney disease, cerebrovascular disease), and findings at presentation (mental status, respiratory rate, blood pressure, temperature, pulse, blood oxygen level, blood pH, sodium, glucose, blood urea nitrogen, hematocrit, and the presence of pleural effusion on a chest X-ray). Each variable contributes a specific number of points to a running total, with age itself carrying heavy weight — a 75-year-old man starts with 75 points before anything else is counted.
The resulting total places a patient into one of five risk classes. Classes I and II correspond to very low 30-day mortality (generally under 1%), class III to low mortality (roughly 1–3%), and classes IV and V to progressively higher risk. In practice, patients in classes I through III are considered “low risk” and are candidates for outpatient treatment, while those in classes IV and V typically need hospital admission. The dividing line between outpatient and inpatient care is the score’s reason for existing.
How It Changes Admission Decisions
Without a structured tool, physicians tend to hospitalize more pneumonia patients than necessary. A multicenter study comparing emergency departments that used the PSI routinely against those that did not found a striking difference: about 43% of low-risk patients were treated as outpatients in PSI-using departments, compared with only about 24% in departments that did not use the score. After adjusting for pneumonia severity, mortality was actually lower among patients treated in PSI-using departments, and other safety outcomes were comparable.2Clinical Infectious Diseases. Routine Use of the Pneumonia Severity Index for Guiding the Site-of-Treatment Decision of Patients with Pneumonia in the Emergency Department: A Multicenter, Prospective, Observational, Controlled Cohort Study
A separate before-and-after study at a single institution found even more dramatic results. After incorporating the PSI into the admission protocol, the admission rate dropped from about 68% to about 30%, while the rate of patients who subsequently needed to be readmitted after initial discharge stayed similar. Direct medical costs per patient fell from roughly $1,500 to roughly $1,200.3PubMed. The effects of incorporating a pneumonia severity index into the admission protocol for community-acquired pneumonia That cost reduction matters because pneumonia is one of the leading reasons for hospital admission worldwide, and every unnecessary inpatient stay carries its own risks — hospital-acquired infections, deconditioning, and the financial burden on patients.
The Infectious Diseases Society of America and the American Thoracic Society jointly recommend preferring the PSI over the competing CURB-65 score for determining hospitalization need, giving the recommendation moderate-to-high quality evidence. Their guidelines state that patients in PSI risk class I or II should be considered candidates for outpatient treatment.4PubMed Central. Infectious Diseases Society of America/American Thoracic Society Consensus Guidelines on the Management of Community-Acquired Pneumonia in Adults
PSI Versus CURB-65
The most common alternative to the PSI is CURB-65, a much simpler tool that scores five items: confusion, elevated blood urea nitrogen, respiratory rate, low blood pressure, and age 65 or older. Because it requires only five data points instead of twenty, CURB-65 is faster to calculate and easier to remember, which has made it popular in settings where speed matters or lab results are not immediately available.
Head-to-head comparisons yield mixed results depending on what you are trying to predict. For 30-day mortality, the PSI and CURB-65 perform in the same ballpark across most studies, though the PSI tends to edge ahead. A large international cohort analysis found that a modified high-risk PSI model had a significantly larger area under the curve than CURB-65 for mortality prediction (0.82 versus 0.77), with better sensitivity across all thresholds tested.5Respiratory Medicine. Comparison of the prognostic performance of the CURB-65 and a modified version of the pneumonia severity index designed to identify high-risk patients using the International Community-Acquired Pneumonia Collaboration Cohort However, an Iranian emergency department study found CURB-65 had higher specificity and overall diagnostic accuracy for mortality prediction than the PSI in that population.6The Brazilian Journal of Infectious Diseases. Pneumonia severity index compared to CURB-65 in predicting the outcome of community acquired pneumonia among patients referred to an Iranian emergency department: a prospective survey
Where the PSI separates itself more clearly is in predicting who needs intensive care. One study found PSI class IV or higher was more accurate than CURB-65 at predicting ICU admission, with an area under the curve of 0.9 versus 0.76.7ERJ Open Research. Scores of community acquired pneumonia for predicting intensive care admission and mortality That said, a systematic review and meta-analysis found that all three major scoring systems — PSI, CURB-65, and CRB-65 — performed similarly when ICU admission was the endpoint, and that tools originally designed to predict mortality performed less well for ICU triage than for death prediction.8PubMed. Severity assessment tools to guide ICU admission in community-acquired pneumonia: systematic review and meta-analysis The takeaway is that neither score was built to decide who needs an ICU bed, and both should be used cautiously for that purpose.
The Age Problem and Elderly Patients
One of the most persistent criticisms of the PSI is that it over-weights age. Because a patient’s age in years directly enters the point total, older adults start with a high baseline score regardless of how they actually look in the emergency department. A vigorous 80-year-old with a mild cough and stable vital signs can land in risk class III or IV simply because of demographics, while a 30-year-old with dangerously low blood pressure might score low enough to be sent home.
Research has confirmed this concern. A study comparing the PSI’s performance across younger adults, elderly patients, and very old patients found that the score underperformed in elderly populations, likely because of the disproportionate weight given to the age variable.9PubMed. Comparison of clinical characteristics and performance of pneumonia severity score and CURB-65 among younger adults, elderly and very old subjects In practical terms, this means the PSI is better at identifying young, low-risk patients who can go home than it is at meaningfully discriminating severity levels among people over 75. CURB-65 also uses age as a binary cutoff (65 or older), but because age contributes only one point out of five rather than a continuous score, the distortion is somewhat less severe.
Why Doctors Frequently Override the Score
Even when the PSI is available and calculated, physicians do not always follow its recommendation. Studies consistently show that a substantial fraction of low-risk patients get admitted anyway. One large observational study found that about 37% of patients classified as low-risk by the PSI were admitted rather than sent home.10PubMed. Reasons why emergency department providers do not rely on the pneumonia severity index to determine the initial site of treatment for patients with pneumonia
The reasons are illuminating. The most common justification for overriding a low-risk PSI score is the presence of a coexisting illness that the score does not fully capture. Over half of patients in one study were admitted because of active conditions like heart failure, angina, or chronic obstructive pulmonary disease — conditions the PSI either does not include or does not weight heavily enough.11JAMA Internal Medicine. Understanding Physician Adherence With a Pneumonia Practice Guideline: Effects of Patient, System, and Physician Factors Other common reasons included deference to the primary care physician’s recommendation for admission, the physician’s own judgment that the pneumonia looked worse than the score indicated, patient preferences, lack of adequate home care support, and previous failure of oral antibiotics.
This pattern reveals something important about how the PSI is meant to work. It was never designed to be the final word. Guidelines describe it as a tool to “inform” the site-of-care decision, not to dictate it. A low-risk score tells the physician it is probably safe to treat this patient at home — but “probably safe” does not account for the patient who lives alone, cannot reliably take medications, or has an unstable housing situation. Clinical judgment fills the gaps the score cannot.
Performance in Immunocompromised Patients
The original PSI was derived from a general population with community-acquired pneumonia, which raises questions about whether it works for people whose immune systems are weakened. The answer depends on the type of immunosuppression. A study dividing immunocompromised patients into two groups found that those with HIV, solid organ transplants, or treatment with immunosuppressive drugs had low in-hospital mortality (about 4%) and could be risk-stratified by the PSI in much the same way as patients with normal immune function. But patients with blood cancers, chemotherapy, chest radiation, or bone marrow transplants had a mortality rate of 20%, and the PSI discriminated poorly among them.12PubMed Central. Pneumonia severity index in the immunocompromised
A broader analysis of healthcare-associated pneumonia confirmed that the PSI was highly sensitive but poorly specific in immunocompromised patients overall, meaning it flagged most high-risk patients but also flagged many who turned out to be fine.13European Respiratory Journal. Severity assessment of healthcare-associated pneumonia and pneumonia in immunosuppression For clinicians, the practical implication is that a high PSI score in someone with a weakened immune system is meaningful, but a low score is less reassuring than it would be in an otherwise healthy person.
Did It Work for COVID-19 Pneumonia?
When COVID-19 overwhelmed hospitals starting in 2020, clinicians needed a way to triage patients quickly, and the PSI was one of the tools pressed into service. Given that the score was developed for bacterial community-acquired pneumonia, its applicability to a novel viral pathogen was not guaranteed.
Multiple studies found that the PSI held up reasonably well. A large multicenter analysis reported that the PSI’s area under the curve for predicting in-hospital mortality was 0.82 in patients with SARS-CoV-2 pneumonia, comparable to the 0.79 it achieved in non-COVID pneumonia patients in the same cohort. The study concluded that the PSI and CURB-65 remained adequate for predicting mortality regardless of whether pneumonia was caused by SARS-CoV-2.14PubMed Central. Pneumonia Severity Index and CURB-65 Score Are Good Predictors of Mortality in Hospitalized Patients With SARS-CoV-2 Community-Acquired Pneumonia Another study found that the PSI alone predicted 30-day mortality in COVID-19 patients significantly better than CURB-65 alone, and that adding C-reactive protein levels to the PSI did not meaningfully improve its performance.15International Journal of Infectious Diseases. Usefulness of pneumonia severity index and CURB-65 in predicting mortality in patients with COVID-19
Earlier work on non-COVID viral pneumonia had already suggested this would be the case. A study examining community-acquired pneumonia in adults with detected respiratory viruses found that the PSI was significantly associated with mortality regardless of whether a virus was present, and that viral pneumonia patients did not have different mortality rates from non-viral patients within the same PSI class.16PLOS ONE. Pneumonia severity index in viral community acquired pneumonia in adults The PSI’s variables capture physiological derangement and baseline health, which matter for prognosis regardless of what caused the infection.
Can Biomarkers Make the Score Better?
Researchers have explored whether adding blood biomarkers to the PSI could sharpen its predictions. C-reactive protein (CRP), procalcitonin, and interleukin-6 are the most commonly studied candidates, since all rise during infection and are available through standard blood tests.
Adding CRP to the PSI pushed the area under the curve from 0.80 to 0.85 in one large study, a statistically significant improvement. However, further adding interleukin-6 or procalcitonin on top of CRP did not improve accuracy any further.17Thorax. Biomarkers improve mortality prediction by prognostic scales in community-acquired pneumonia A separate study specifically evaluating procalcitonin found that simply appending it to the PSI score produced only a minimal bump, with sensitivity increasing from about 72% to 76% and specificity from about 77% to 80%.18Annals of Emergency Medicine. Risk Prediction With Procalcitonin and Clinical Rules in Community-Acquired Pneumonia
The practical question is whether a 5-point improvement in AUC justifies the added cost and delay of ordering extra labs. For a busy emergency department where the PSI already requires gathering 20 variables, the marginal value of one more blood marker is debatable. CRP is cheap and widely available, which gives it an edge over procalcitonin, but guidelines have not yet formally incorporated any biomarker into the standard PSI calculation.
Long-Term Mortality and the Score’s Blind Spot
The PSI was designed to predict 30-day mortality, and that is what it does best. But pneumonia, especially in older adults, often triggers a cascade of complications — deconditioning, worsening of chronic diseases, rehospitalization — that extends well beyond the first month. A study of older adults with community-acquired pneumonia found that the PSI was associated with long-term mortality in univariate analysis, alongside mechanical ventilation, mental status changes, presence of cancer, and hospital readmission.19Journal of Infection in Developing Countries. Predictors of Long-term Outcomes in the Older Adults with Community-Acquired Pneumonia
However, when evaluated as a standalone predictor of one-year mortality, the PSI’s performance drops off. In a study of nearly 7,000 hospitalized pneumonia patients, one in four died within a year of discharge. The PSI’s area under the curve for that endpoint was only 0.72, and the study’s authors characterized both the PSI and CURB-65 as weak predictors of long-term mortality.20Norton Healthcare Medical Journal. Pneumonia Severity Index And CURB-65 Scores Are Poor Predictors For Long-Term Mortality In Hospitalized Patients With Community-Acquired Pneumonia This makes sense when you consider what the PSI includes and excludes. It captures acute physiological disturbance and a few broad comorbidity categories, but it does not capture frailty, functional status, nutritional state, or the number and severity of chronic conditions in any granular way. Those factors matter much more for long-term survival than for surviving the next month.
Challenges in Resource-Limited Settings
The PSI’s 20 variables assume access to laboratory testing, arterial blood gas analysis, and chest imaging at the point of care — infrastructure that is standard in high-income emergency departments but often unavailable in low- and middle-income countries. A review of pneumonia severity scores in resource-poor settings raised concerns about the generalizability of scores developed in well-equipped health systems, noting that laboratory values making up a large portion of the PSI may simply not be obtainable in many facilities.21PubMed Central. Pneumonia severity scores in resource poor settings
In these settings, CURB-65 or even its simplified variant CRB-65 (which drops blood urea nitrogen entirely) may be more practical, since they can be calculated at the bedside with minimal equipment. The trade-off is some loss in predictive accuracy, but a score that can actually be completed in the field is more useful than a more accurate one that cannot. There have been calls for new severity tools specifically derived from populations in lower-resource regions, where the epidemiology of pneumonia, the prevalence of HIV and tuberculosis co-infection, and the age distribution of patients can differ substantially from the North American and European cohorts the PSI was built on.
Electronic Health Records and Automated Scoring
One way to address the PSI’s complexity is to let the computer calculate it. A study examining whether clinical data routinely available in a computerized patient record system could reliably drive the PSI found that the data quality was sufficient to automate the score and present it to physicians in real time at the point of care.22Journal of the American Medical Informatics Association. Assessing the Quality of Clinical Data in a Computer-based Record for Calculating the Pneumonia Severity Index Modern electronic health record systems can pull the 20 required variables from labs, vital signs, and demographics that are already being entered, calculate the risk class, and display it without requiring the physician to look anything up or do any arithmetic.
Automated PSI calculation removes one of the score’s biggest practical barriers. When clinicians have to calculate it by hand or through a separate app, many simply skip it and rely on gut feeling. When it appears automatically alongside the patient’s chart, it becomes ambient decision support rather than an extra task. Hospitals that have integrated automatic PSI scoring into their emergency department workflows have generally reported higher rates of appropriate outpatient treatment for low-risk patients, which aligns with what the tool was designed to achieve in the first place.