How Pneumonia Severity Scores Help Decide Who Needs the Hospital

A pneumonia severity score tells you how likely a patient is to die within 30 days. It does not tell you whether that patient can safely go home. Holding those two ideas apart is the single most useful thing to understand about the Pneumonia Severity Index (PSI) and CURB-65, the two tools most clinicians reach for when someone arrives with community-acquired pneumonia and the first real question is where care should happen: at home, on a hospital ward, or in the intensive care unit.

Key points#

• PSI and CURB-65 were built to predict 30-day mortality, a prognostic question, not a disposition decision.
• A low predicted risk of death is a reason to consider home care, not a rule that home care is safe.
• Reasons to admit that no score captures include low oxygen, an inability to keep down medicine, an unstable chronic illness, and no support at home.
• The 2019 ATS/IDSA guideline recommends a validated rule, preferentially the PSI, used together with clinical judgment.

Two ways to ask "how sick is this person"#

There is a gap between two questions that sound almost identical at the bedside. The first is prognostic: over the next month, how likely is this person to die? The second is practical: right now, does anything about this person's condition or circumstances mean home cannot deliver safe care? PSI and CURB-65 answer the first question well. They were never designed to answer the second, and most of the trouble with these scores comes from treating a good answer to the first as if it settled the second.

What each score counts#

The Pneumonia Severity Index, published by Fine and colleagues in 1997, is the more detailed of the two. It sorts adults into five risk classes using roughly twenty variables: age, sex, whether the person lives in a nursing home, coexisting illnesses such as cancer, liver disease, heart failure, and kidney disease, and a spread of vital signs and laboratory values including blood pressure, respiratory rate, oxygenation, blood urea nitrogen, sodium, and arterial pH. Classes I through III carry a low 30-day mortality, class IV an intermediate risk, and class V the highest. Because it was derived from more than fourteen thousand inpatients and validated in tens of thousands more, guidelines treat it as a reference standard.

CURB-65, derived by Lim and colleagues in 2003, trades that detail for speed. It gives one point each for Confusion, elevated blood Urea, a Respiratory rate of 30 or more, low Blood pressure, and age 65 or older. A total of 0 or 1 marks a low-risk group, 2 an intermediate group, and 3 to 5 a high-risk group. A clinician can add it up at the bedside without waiting for a full metabolic panel, which is why it is popular in emergency departments. An even shorter version, CRB-65, drops the urea so the score needs no bloodwork at all.

Both are genuinely useful. Both discriminate mortality risk far better than an unaided guess. Neither knows anything about the ward.

Where the number and the decision come apart#

Two short cases show the seam. Picture a healthy 30-year-old with pneumonia whose oxygen saturation is 88 percent. Age drives so much of the PSI that this patient can fall into a low-risk class, yet that degree of low oxygen usually calls for supplemental oxygen and monitoring a home cannot provide. The score sees a small chance of dying; the clinician sees someone who needs admission today.

Now the reverse. An older adult with a high-looking score may in fact be stable, already improving on treatment, and better off at home with support than admitted to a ward, where the risks of delirium and hospital-acquired infection are real. The number points one way and the person points the other.

Death is only one of several reasons to keep someone in the hospital. An inability to swallow or keep down oral antibiotics and fluids, pain that is not controlled, a chronic illness that is decompensating, blood pressure that will not hold, or simply no one at home able to help can each justify admission on their own, regardless of a reassuring class or a CURB-65 of 1. A score cannot see any of that. Treating it as the entire answer is a category error, not a rounding error.

What the 2019 ATS/IDSA guideline actually says#

The 2019 community-acquired pneumonia guideline from the American Thoracic Society and the Infectious Diseases Society of America is often quoted loosely, so the exact shape of its recommendation matters. It makes a strong recommendation, on moderate-quality evidence, to determine the site of care using a validated clinical prediction rule for prognosis, preferring the PSI over CURB-65. The reason given is that the PSI classifies a larger share of patients as low risk and discriminates mortality more sharply than CURB-65 does.

Two conditions attached to that recommendation carry as much weight as the recommendation itself. First, the rule is meant to be used together with clinical judgment rather than instead of it, precisely because a prognostic score misses the admission drivers described above. Second, the panel advises starting empiric antibiotics in adults with confirmed pneumonia no matter what the initial serum procalcitonin level is, so that biomarker is not the tool for deciding whether to treat. For the sickest patients, the guideline turns to a separate set of criteria for severe pneumonia and direct ICU admission, such as septic shock needing vasopressors or respiratory failure needing mechanical ventilation, rather than to any PSI class or CURB-65 total.

Using a score without over-trusting it#

The most defensible habit is to treat a low score as a floor, not a verdict. It grants permission to consider outpatient care; it never commands a discharge. A clinician can and should revise upward when oxygenation, oral intake, comorbidity, or home circumstances demand it. These tools are least reliable at exactly the borderline cases where they get used the most, which is the whole reason the guideline pairs them with judgment instead of handing the decision over to them.