The lung's job is to put a very large surface in contact with air. It does that with a very large number of alveoli — thin-walled sacs, each wrapped in capillaries, where oxygen crosses into the blood and carbon dioxide crosses out. The barrier between air and blood is about a micrometre thick, because gas exchange needs it to be.
Pneumonia is infection of those air sacs. Bacteria, viruses or fungi reach the alveoli and the lung mounts an inflammatory response: neutrophils and fluid pour into spaces that are supposed to contain nothing but air. The consequence is mechanical before it is anything else. Alveoli full of fluid and pus cannot exchange gas. Blood flowing past them returns to the heart still deoxygenated. That is why the cardinal symptoms are cough, fever, breathlessness and pleuritic chest pain — pain that is sharp and worse on breathing in, because the inflammation has reached the pleural surface — and why the cardinal sign on a chest radiograph is an infiltrate: the fluid-filled region shows up white where air would be black.
It is one of the world's leading infectious killers. The Global Burden of Disease study estimated that lower respiratory infections caused 2,377,697 deaths in 2016 across all ages — including 652,572 in children under 5 and 1,080,958 in adults over 70 — with Streptococcus pneumoniae the single leading cause, contributing more deaths than all other aetiologies combined at 1,189,937 [1].
Those numbers contain a correction to the popular picture, and the paper never states it. Pneumonia is widely framed as a childhood killer. Adding the two age extremes, the over-70s account for 1.66 times as many lower-respiratory-infection deaths as the under-5s, and together the two groups are 72.9 percent of the global total. Both ends of life are vulnerable; the older end is now the larger burden.
Severity spans almost the entire clinical range. The same diagnosis covers a person treated at home with tablets and someone ventilated in intensive care, and the difference is not a different disease but a different position on a risk gradient. In the derivation cohort of the score this review is built around, 30-day mortality ran from 0.7 percent to 57 percent depending on five findings available at the bedside [2]. Deciding where a patient should be treated is therefore the first and most consequential clinical act, and it is what the centerpiece is about.
And a great deal of it is preventable. Vaccines against the leading bacterial cause and against the viruses that precede it work, and are covered below. So is the most basic finding in the paediatric data: childhood wasting remains the leading risk factor for under-5 lower-respiratory-infection mortality, responsible for 61.4 percent of those deaths [1]. In the highest-burden countries, improving wasting, household air pollution, ambient particulate pollution and antibiotic access could avert one under-5 death for every 4,000 children treated [1]. The biggest available gains are not pharmaceutical.
Centerpiece: five bedside variables and the disposition decision
A patient arrives short of breath with a fever and an infiltrate on the chest film. The question that has to be answered in the next few minutes is not which organism it is — that will take days, if it is ever established — but where this person should be treated. Home, a ward bed, or intensive care. Get it wrong in one direction and someone deteriorates unmonitored at home; get it wrong in the other and a hospital fills with people who did not need admitting.
CURB-65 answers it with five findings, each worth one point: Confusion, blood Urea above 7 mmol/l, Respiratory rate of 30 or more per minute, low Blood pressure (systolic below 90 or diastolic 60 or less), and age 65 or over [2]. It was derived and validated across three prospective cohorts in the UK, New Zealand and the Netherlands — 1,068 patients, mean age 64, overall 30-day mortality 9 percent — using multiple logistic regression with 30-day mortality as the outcome [2].
The resulting gradient is the left panel: 0.7 percent at score 0, 3.2 percent at 1, 3 percent at 2, 17 percent at 3, 41.5 percent at 4, and 57 percent at 5 [2].
What the score buys, stated as a number neither paper gives. That range is an 81-fold spread in 30-day mortality, produced by five things a clinician can establish with a conversation, a blood pressure cuff, a watch and one blood test — in a cohort whose overall mortality was 9 percent. Sorting an undifferentiated group of pneumonia patients into strata running from well below to far above the average, without imaging, culture or any specialised equipment, is the entire case for bedside risk scoring.
And there is a kink in the original series that nobody remarks on. Mortality at score 1 is 3.2 percent and at score 2 is 3.0 percent. The score goes up and the risk goes very slightly down — the series is not monotonic at exactly the point where the outpatient/admit boundary sits.
An independent cohort resolves it. In a consecutive series at a 400-bed teaching hospital — 1,100 inpatients plus 676 outpatients, 1,776 in total — 30-day mortality by CURB-65 score was 0 percent, 1.1, 7.6, 21, 41.9 and 60 percent [3]. That series is monotonic throughout, so the kink in the original is a small-sample artefact, which is what an independent validation is for. The same study confirmed the score's association with need for mechanical ventilation, admission rate and length of stay.
The check the arithmetic was never given, and it lands on the decision boundary. Both papers print a table; nobody puts the two side by side. Doing so shows that the two cohorts differ by an average of 2.47 percentage points at the top of the score (3–5) and by an average of 2.47 percentage points at the bottom (0–2) — identical to four decimal places, which is a coincidence, and a useful one because it isolates what changes.
Those same 2.47 points are 6 percent of the value at the top of the score and 95 percent of it at the bottom — a fifteen-fold worse relative agreement precisely where the send-home decision is made. At a score of 4 the two cohorts say 41.5 and 41.9 percent, and no clinical decision turns on the difference. At a score of 2 they say 3.0 and 7.6 percent, and the difference is between an outpatient and an inpatient. The score is most reliable exactly where it is least needed, and least reliable where it decides the most.
The other score was built for the other end. The Pneumonia Severity Index takes a different approach — twenty variables spanning age, coexisting disease, physical findings and laboratory results, sorted into five risk classes — and was derived on 14,199 inpatients and validated on 38,039 inpatients plus 2,287 patients in the PORT cohort, a derivation cohort thirteen times the size of CURB-65's. Its design goal was explicitly to identify low-risk patients, and it does: mortality was 0.1–0.4 percent in class I, 0.6–0.7 percent in class II and 0.9–2.8 percent in class III, with only seven deaths among the 1,575 PORT patients in the three lowest classes — and only four of those pneumonia-related [4] [5] [6].
Head-to-head, the two have different strengths, and the honest comparison says so. In a prospective cohort of 392 emergency-department patients, CURB-65 predicted mortality well but performed less well when need for ICU admission was included in the outcome, while the PSI and CURB performed similarly across outcomes and the modified BTS score did best overall [7]. The severity-prediction rules have been validated repeatedly with mixed results — one UK validation found the BTS rule 52 percent sensitive and the modified rule 66 percent sensitive for death, notably worse than in the derivation studies, in a population with a high proportion of very elderly patients [8] — and further rules have been developed specifically to identify severe pneumonia rather than death, one reaching an area under the curve of 0.92 and outperforming CURB-65, the PSI and the modified ATS rule for that particular target [9]. Prognostic factors have been characterised independently across many cohorts [10] [11] [12] [13] [14] [15] [16], including alcohol intake as both a risk and a prognostic factor [17], and radiographic findings at presentation have been tested as mortality predictors [18].
Two honest limits on the figure. The three disposition bands drawn on the left panel are the conventional reading of the score; the founding paper states only that it stratifies patients "into different management groups", so the specific cut-points are illustrative rather than quoted. And CURB-65 predicts mortality, which is not the same as predicting need for a ward bed — a young person with hypoxia and no CURB-65 points may still need admitting, and the score does not know that.
The teaching point. Five variables, obtainable in minutes, spread 30-day mortality across nearly two orders of magnitude — enough to decide where someone is treated. That is risk-based triage working. The right panel is the caveat that comes with it: the strata at the bottom of the score, where the score is used to send people home, are exactly the strata on which independent cohorts agree least.
Pillar 1: measurement and diagnosis
The chest radiograph, and how much it actually settles
An infiltrate on the chest film is the standard confirmation of pneumonia, and it is genuinely necessary — clinical features alone are unreliable enough that prediction rules exist purely to decide who needs the film [19] [20]. But three limitations deserve stating.
Observers disagree. Interobserver reliability of the chest radiograph in community-acquired pneumonia has been formally measured [21], and standardised interpretation protocols had to be developed for paediatric epidemiological studies precisely because unstandardised reading was not reproducible enough to count cases [22] [23].
It cannot tell you the organism. Five observers reading 36 chest films from patients with laboratory-proven diagnoses achieved a sensitivity for bacterial pneumonia of 42 to 58 percent [24]. Radiographic features of Legionnaires' disease, pneumococcal pneumonia and mycoplasma have been compared directly and overlap substantially [25] [26]. The film tells you there is pneumonia, not what caused it.
And it can be normal early. Patients admitted with suspected pneumonia and a normal chest radiograph are a described population with their own epidemiology and microbiology [27], high-resolution CT detects infiltrates the plain film misses [28] [29], and in mechanically ventilated patients the diagnostic accuracy of the portable film against autopsy findings is poor [30]. Radiographic clearing also lags recovery by weeks [31], which is why a persisting shadow in someone who feels well is usually not treatment failure.
The categories, which change everything downstream
Pneumonia is classified by where it was acquired, because that predicts the organism and therefore the antibiotic.
Community-acquired pneumonia (CAP) is the ordinary case, managed under long-standing consensus guidelines [32] [33].
Hospital-acquired and ventilator-associated pneumonia (HAP/VAP) arise in patients already in hospital, and the organisms are different and more resistant [34] [35]. Ventilator-associated pneumonia is common: 177 of 1,014 mechanically ventilated patients (17.5 percent) developed it at a median 7 days, with the daily hazard falling after day 5 — 3.3 percent per day at day 5, 2.3 percent at day 10, 1.3 percent at day 15 — and independent risk factors including burns, trauma, central nervous system disease, witnessed aspiration and paralytic agents [36]. Incidence, attributable mortality and cost have been quantified repeatedly [37] [38] [39] [40] [41], along with the debates the field has never fully settled [42] [43].
A cautionary tale about a third category. "Healthcare-associated pneumonia" was introduced to capture patients from nursing homes and dialysis units who were thought to carry hospital-type organisms [44], and it was written into guidelines [34]. It was subsequently dropped, because it drove broad-spectrum treatment in patients who mostly did not need it. That reversal is a useful reminder that a category invented to improve empirical therapy can make it worse.
Aspiration — inhalation of oropharyngeal or gastric contents — is the fourth route, and witnessed aspiration is among the strongest independent risk factors for ventilator-associated pneumonia [36].
Typical, atypical, and viral
Typical bacterial pneumonia is dominated by Streptococcus pneumoniae, still the leading global cause of lower-respiratory-infection death [1] [45] [46], with Haemophilus influenzae and Moraxella also contributing.
Atypical pathogens — Mycoplasma pneumoniae, Chlamydia pneumoniae and Legionella — are so named because they do not respond to beta-lactams, which act on a cell wall that mycoplasmas do not have and that the intracellular organisms are shielded from. Mycoplasma pneumoniae causes more illness than its reputation suggests, with extrapulmonary complications across multiple organ systems and severity apparently tracking the vigour of the host response [47]. Chlamydia pneumoniae was identified as the TWAR strain in the 1980s [48] [49] and quantified in prospective series [50] [51]. Atypicals are common everywhere: across 4,337 patients in four world regions, their incidence in CAP was 20 to 28 percent [52]. Legionella in particular punishes delay — delayed appropriate therapy is associated with increased mortality [53] — and has emerged as a specific hazard in patients on TNF antagonists [54].
Viruses are the commonest finding when you look properly, and this is the single most surprising result in modern CAP epidemiology. In the EPIC study, 2,320 US adults hospitalised with radiographically confirmed CAP underwent systematic blood, urine and respiratory testing by culture, serology, antigen detection and molecular methods [55]. A pathogen was detected in only 38 percent — one or more viruses in 23 percent, bacteria in 11 percent, both in 3 percent. The commonest single agents were human rhinovirus (9 percent), influenza (6 percent) and S. pneumoniae (5 percent) [55]. In the paediatric arm, detection was much better — a pathogen in 81 percent of 2,222 children, viruses in 66 percent and bacteria in 8 percent, with RSV in 37 percent of under-5s and Mycoplasma in 19 percent of those 5 and over [56].
Two conclusions follow, and both matter. Respiratory viruses cause more hospitalised pneumonia than bacteria do. And in the majority of adults, no pathogen is identified at all, despite comprehensive modern testing. Empirical antibiotic therapy is not a failure of diligence; it is the normal condition of this disease.
Microbiology, and its modest yield
Sputum Gram stain is quick and specific but limited by sample quality: in 533 hospitalised patients only 210 (39 percent) produced a good-quality sample, of which 175 showed a predominant morphotype, giving a sensitivity of 57 percent and specificity of 97 percent for pneumococcal pneumonia [57]. Blood cultures are positive in a minority, and their usefulness rises with severity [58]. Pneumococcal urinary antigen is the mirror image of the Gram stain — sensitivity 65.9 percent, specificity 100 percent in 220 adults [59] [60] — so a positive result is decisive and a negative one is uninformative. Multiplex PCR improved on all of them: a combined real-time assay was significantly more sensitive than conventional methods for the main viruses and atypical bacteria, and returned results within a clinically relevant window [61] [62].
In ventilated patients, sampling is invasive and the debate over technique is old and largely settled toward pragmatism. Quantitative cultures of bronchoalveolar lavage, protected specimen brush and endotracheal aspirates have been compared extensively [63] [64] [65] [66] [67] [68] [69] [70], invasive quantitative sampling was tested against clinical management and found to affect outcome [71], but the counter-position that invasive testing is not routinely needed was argued early [72], and a later comparison of quantitative against qualitative cultures found no clinical-outcome advantage [73].
Biomarkers
Procalcitonin rises in bacterial infection and much less in viral illness [74] [75], and the diagnostic literature comparing it with C-reactive protein is large [76] [77] [78]. Its real value has turned out to be not diagnosis but stopping antibiotics, which is covered under stewardship below. Procalcitonin kinetics also carry prognostic information in ventilator-associated pneumonia [79].
Pillar 2: treatment and prevention
Empirical antibiotics: choose by setting, severity and risk
Because the organism is usually unknown at the moment of treatment — and in most adults will never be known [55] — initial therapy is empirical, chosen to cover what is likely given where the pneumonia was acquired and how sick the patient is. Following the guideline matters: among 54,619 adults, initial guideline-concordant therapy was associated with lower in-hospital mortality (odds ratio 0.70), less sepsis and renal failure, and about 0.6 fewer days of stay and of parenteral therapy [80] [81].
For non-severe CAP, a beta-lactam is the backbone, and the atypical question is genuinely unsettled. A meta-analysis of trials in non-severe CAP found no advantage for antibiotics active against atypical pathogens over beta-lactams overall (relative risk 0.97) — with one exception, patients later shown to have Legionella, in whom atypical coverage was clearly better (0.40, 0.19–0.85) — and concluded that beta-lactams should remain the initial choice [82]. A Cochrane review of 28 trials and 5,939 hospitalised patients likewise found no mortality difference between regimens with and without atypical cover, with an apparent bacteriological advantage that disappeared when only high-quality studies were analysed [83].
But for hospitalised and severe pneumonia the balance shifts the other way. Across 4,337 patients in four regions, those treated with atypical coverage had shorter time to clinical stability (3.2 versus 3.7 days), shorter stay (6.1 versus 7.1 days) and lower total mortality (7 versus 11.1 percent) [52]. Adding a macrolide to a beta-lactam in bacteraemic pneumococcal pneumonia was associated with lower in-hospital mortality [84], monotherapy appeared suboptimal for severe bacteraemic pneumococcal disease [85], and macrolide combination improved survival in intubated CAP patients [86]. A randomised non-inferiority trial then failed to show that beta-lactam monotherapy was non-inferior to beta-lactam plus macrolide in moderately severe CAP: 41.2 percent versus 33.6 percent had not reached clinical stability at day 7, with the upper confidence limit exceeding the non-inferiority boundary, and the deficit concentrated in patients infected with atypical pathogens (hazard ratio 0.33) or with PSI class IV disease — while those without atypical infection or with PSI I–III did equally well on monotherapy [87]. Part of the macrolide effect may not be antibacterial at all: their anti-inflammatory properties are a recognised and underappreciated contributor [88].
That is a coherent picture rather than a contradiction. The sicker the patient and the more likely an atypical organism, the more atypical coverage is worth — and in mild outpatient disease it mostly is not.
Resistance constrains the choices. Macrolide resistance in S. pneumoniae is well characterised mechanistically [89] [90] and clinically consequential: macrolide treatment failure occurs in bacteraemia due to erythromycin-resistant pneumococci [91].
For HAP and VAP the spectrum is broader, because the organisms include Pseudomonas aeruginosa, Acinetobacter baumannii and MRSA [34] [92] [93] [94] [95]. Multidrug resistance in VAP pathogens has measurable impact on outcome [96], appropriate initial therapy matters greatly [97] [98], and combination versus monotherapy has been tested directly [99] [100] [101]. Inhaled antibiotics have been used as adjuncts for resistant Gram-negative VAP [102] [103].
De-escalation, short courses and stopping
This is where the field has moved most, and the direction is consistently toward less.
Short courses work. The landmark trial randomised patients with VAP who had received appropriate initial therapy to 8 versus 15 days: there was no excess mortality (18.8 versus 17.2 percent) and no more recurrent infection (28.9 versus 26.0 percent), but 4.4 more antibiotic-free days. The exception was VAP caused by non-fermenting Gram-negative bacilli including Pseudomonas, where recurrence was higher on the short course (40.6 versus 25.4 percent). And among patients who did have recurrences, multiresistant pathogens emerged less often after 8 days than after 15 (42.1 versus 62.0 percent) [104] — the short course was not merely equivalent, it left behind less resistance. A Cochrane review of six studies and 1,088 participants confirmed the pattern: short courses increased antibiotic-free days and reduced recurrence due to multi-resistant organisms (odds ratio 0.44) without harming mortality [105] [106] [107]. Short-course strategies have also been tested in CAP [108] [109] and in undifferentiated pulmonary infiltrates in the ICU [110].
De-escalation — narrowing therapy once cultures return — is established practice in VAP [111] [112], and a randomised antibiotic-discontinuation policy has been tested for clinically suspected VAP [113].
Procalcitonin can tell you when to stop. In ProHOSP, 1,359 patients with lower respiratory tract infections randomised to procalcitonin guidance versus standard guidelines had similar adverse outcomes (15.4 versus 18.9 percent) but mean antibiotic exposure of 5.7 versus 8.7 days (a 34.8 percent reduction) — with the reduction largest in acute bronchitis (1.0 versus 2.8 days) and substantial in CAP (7.2 versus 10.7 days) — and fewer antibiotic-associated adverse effects (19.8 versus 28.1 percent) [114]. A Cochrane review across 32 randomised trials reached compatible conclusions [115] [116] [117], as did procalcitonin-guided strategies in the ICU [118] [119] [120] [121] and in CAP specifically [122] [123].
The stewardship point, stated plainly. Respiratory viruses cause more hospitalised adult pneumonia than bacteria do [55] [56], and antibiotics do nothing for them. But no available test reliably distinguishes viral from bacterial pneumonia at the moment treatment must start — the radiograph cannot [24], and procalcitonin shifts probabilities rather than settling them [76]. The practical resolution is not to withhold antibiotics from sick patients on a guess, but to start them and then stop early, guided by clinical stability [124] and, where available, by procalcitonin [114] [115]. Antimicrobial resistance is a global multifaceted problem [125] [95], and duration is the part of it that individual prescribers most directly control.
Supportive care, and knowing when someone is better
Oxygen, fluids and, where needed, ventilatory support are the substrate on which antibiotics work. Judging recovery has been formalised: explicit criteria for time to clinical stability in hospitalised CAP found a median of 2 days to stability by the most lenient definition and 7 by the most conservative, with sicker patients taking longer — and, importantly, once stability was reached, deterioration requiring intensive care occurred in 1 percent of cases or fewer [124]. The same study found substantial inefficiency: 65 to 86 percent of patients stayed more than a day after reaching stability, and fewer than 29 to 46 percent, depending on the stability definition used, were switched to oral antibiotics within a day of it [124]. Those criteria are the evidence base for switching to oral therapy and discharging.
Prevention
Pneumococcal conjugate vaccines are the major success. The heptavalent conjugate was 97.4 percent efficacious (95% CI 82.7–99.9) against invasive disease caused by vaccine serotypes in fully vaccinated children, and 93.9 percent in intention-to-treat, with additional effects on otitis media [126] [127]. A nine-valent formulation worked in children with and without HIV infection [128] [129]. Effectiveness studies confirmed the trial results in practice [130] [131], pneumonia admissions fell after routine childhood immunisation [132], and drug-resistant pneumococcal disease declined as well [133].
The mechanism includes an indirect effect that is arguably the more important one. Conjugate vaccines reduce nasopharyngeal carriage [134] [135] [136], so vaccinated children stop transmitting. That produces herd protection in adults who were never vaccinated: among US adults aged 50 and over, invasive pneumococcal disease fell 28 percent between 1998–99 and 2002–03, driven by a 55 percent fall in disease caused by the seven conjugate serotypes — in a population being given a different vaccine [137]. A childhood vaccine cut disease in their grandparents.
The honest counterweight is serotype replacement. Pneumococci not covered by the vaccine expand into the vacated niche. Incidence of disease due to non-PCV7 serotypes rose in the United States [138], invasive disease from non-vaccine serotypes appeared among Alaska Native children with high vaccine coverage [139], and the England and Wales analysis quantifies both sides at once: after eight years of conjugate vaccine use, overall invasive pneumococcal disease was down 56 percent against the pre-PCV7 baseline (15.63 to 6.85 per 100,000), with an 86 percent reduction in PCV7 serotypes and 69 percent in the additional PCV13 serotypes — while non-PCV13 serotypes increased by 25 percent [140] [141]. The net effect is strongly favourable and the replacement is real, which is why valency keeps rising.
The polysaccharide vaccine is the older adult product, with efficacy of 57 percent overall against invasive disease and 75 percent in immunocompetent people over 65, not declining measurably with time since vaccination [142] [143]. Recommendations now combine conjugate and polysaccharide vaccines in adults and in immunocompromised groups [144] [145], with immunogenicity of the 13-valent conjugate established in adults 70 and over [146] [147] and WHO position articulated for childhood immunisation [148].
Influenza vaccination prevents pneumonia too, both directly and by removing the viral insult that precedes bacterial superinfection. Efficacy in elderly persons is established [149] [150] [151] [152], hospital admissions for pneumonia in non-institutionalised elderly people fell as a result of vaccination [153], and mortality among adults hospitalised with CAP was lower in those vaccinated [154] — including, strikingly, outside the influenza season [155], which raises the question of confounding by healthy-user effects as much as it demonstrates benefit. Influenza and pneumococcal vaccines can be co-administered [156] [157] [158] [159], vaccinating healthcare workers who care for the elderly has been assessed [160], and vaccination in pregnancy is recommended [161].
Smoking cessation removes the single most important modifiable host defence impairment in adults, and in the paediatric setting the corresponding targets are wasting and household air pollution [1].
The biology, briefly
The alveolus is not passive. Resident alveolar macrophages and the epithelium detect conserved microbial structures through pattern-recognition receptors, and two of these are directly implicated in pneumonia. Toll-like receptor 4 recognises pneumolysin, the pneumococcal toxin, and that recognition confers resistance to pneumococcal infection [162]; TLR4's role has been dissected in both Gram-positive and Gram-negative pneumonia in mice [163], where in E. coli pneumonia it mediates the inflammatory response but not bacterial elimination [164] — a dissociation that explains why blocking inflammation in pneumonia has been so difficult to turn into therapy. Toll-like receptor 2 contributes to the early inflammatory response in murine pneumococcal pneumonia [165] and drives inflammasome activation downstream [166] [167]. Neutrophil recruitment depends on IL-23-driven IL-17 production in lung defence [168], iron sequestration by lipocalin 2 is required for defence against Klebsiella — and is evaded by bacteria carrying the iroA gene cluster [169] [170] — and prior bacterial colonisation can dampen influenza-mediated lung injury by inducing M2 alveolar macrophages [171]. The recurring theme is that most of what makes pneumonia dangerous is the host response, not the organism.
Pillar 3: progress
Rapid molecular diagnostics coupled to stewardship. Multiplex PCR detects viruses and atypical bacteria faster and more sensitively than culture and serology [61] [62], and the EPIC results show what it reveals: viruses predominate and most adult cases have no identified pathogen at all [55] [56]. The value of a faster answer is realised only if it changes prescribing, which is why diagnostics and stewardship have to advance together [125].
Higher-valency conjugate vaccines, driven by the serotype replacement the earlier ones caused [140] [138] [139], with combined conjugate-plus-polysaccharide schedules in adults [144] [145].
Shorter courses, now supported by randomised evidence in VAP [104] [105] and CAP [108], with the specific and important finding that shorter treatment leaves behind less resistance among recurrences, not more [104].
Host-response biomarkers to decide when to stop rather than what to start [114] [115] [116], with procalcitonin the only one with randomised evidence at scale and its limitations well described [172] [173].
A gap this substrate covers unevenly. COVID-19 arrived in this literature as "pneumonia of unknown aetiology" [174] [175] [176], and the substrate contains the early Wuhan clinical series in quantity [177] [178] [179] [180] [181] [182] [183] alongside molecular diagnosis [184] and co-infection [185]. What it does not contain is the COVID-19 vaccine literature. The task specification lists COVID vaccination alongside pneumococcal and influenza vaccines as prevention; this review cannot ground that claim from this substrate and therefore does not make it. The same applies to lung ultrasound, which the substrate does not cover despite its growing use.
Dig deeper in lmmol
COPD is the commonest chronic lung disease predisposing to pneumonia, and the two intersect at the bedside constantly — bacterial airway load rises during exacerbations [186], and procalcitonin-guided stewardship produced its largest antibiotic reduction in exacerbations of COPD, cutting exposure by 50.4 percent [114]. Tuberculosis is the other great bacterial lung infection and the instructive contrast: a slow granulomatous disease requiring months of multi-drug therapy, against an acute alveolar one treated in days — different organisms, different tempo, same organ, and a chest radiograph that can look deceptively similar. HIV/AIDS transformed pneumonia into an opportunistic problem and generated its own literature on Pneumocystis prophylaxis and when it can safely be stopped after immune reconstitution [187] [188] [189]; HIV also reshaped pneumococcal epidemiology, with the HIV-infected proportion of invasive pneumococcal cases rising from 1.7 to 5.6 percent over five years in US surveillance [137], and the nine-valent conjugate vaccine was tested specifically in children with and without HIV [128]. H5N1 avian influenza is the pandemic-potential version of the viral pneumonia this review describes, and the connection to ordinary practice is direct: influenza is the second most commonly detected pathogen in US adults hospitalised with CAP [55] and influenza vaccination reduces pneumonia admissions [153]. Sepsis is what severe pneumonia becomes — pneumonia is its commonest source, guideline-concordant antibiotic therapy reduced sepsis as an outcome [80], and the procalcitonin literature is shared between the two fields [190] [191]. Asthma shares the atypical-pathogen literature, with Mycoplasma and Chlamydia pneumoniae implicated in both children and adults [192] [193]. On the molecular side, Toll-like receptor 4 and Toll-like receptor 2 are the receptors through which the alveolus detects the organisms in this review [162] [165] [163]. The full collection is at health.