Do symptoms and initial clinical probability predict the radiological diagnosis of acute sinusitis in children?

Abstract

OBJECTIVE:

To evaluate the value of signs and symptoms in children for the radiological diagnosis of acute sinusitis.

DESIGN:

Prospective cohort study.

SETTING:

University-affiliated tertiary care hospital.

PATIENTS:

All children presenting with symptoms suggestive of acute sinusitis for whom sinus radiographs were ordered.

METHODS:

Data were collected on the presence of specific symptoms and the initial probability of sinusitis. Criterion-based radiological diagnoses were made.

RESULTS:

Three hundred ninety-two consecutive children were seen; 257 children had a radiological diagnosis of acute sinusitis (66%), 128 patients (33%) presented with complete opacity of at least one sinus and 14 (4%) children had an air-fluid level. Sensitivity, specificity, predictive values and likelihood ratios were measured for clinical findings. Classical symptoms (rhinorrhea lasting more than 10 days and purulent rhinorrhea) increased the likelihood ratios the most (1.3 and 1.34, respectively). Logistic regression showed two independent predictors: purulent rhinorrhea (odds ratio 2.0) and the presence of acute otitis media (odds ratio 2.6). The initial clinical probability was more accurate than any other single finding: high probability (likelihood ratio 2.0), intermediate probability (likelihood ratio 1.1) and low probability (likelihood ratio 0.6).

CONCLUSION:

Classical symptoms are predictive of the presence of acute sinusitis as diagnosed on sinus radiographs. The physician’s overall clinical impression, expressed as an initial probability, was superior to any single historical or examination finding in the diagnosis of acute sinusitis.

Sinusitis is common, complicating 5% to 10% of upper respiratory tract infections (1). In adults and adolescents, the symptoms of sinusitis include nasal obstruction, nasal discharge, facial pain, headache and fever (2). In children, complaints are less specific and have considerable overlap with those of a common cold (1).

Sinus radiographs are considered to be the standard for diagnosing sinusitis. The presence of an air-fluid level within the maxillary sinus or maxillary opacification in a patient with sinusitis symptoms correlates with the recovery of pathogenic bacteria from maxillary sinus aspiration in more than 70% of children (3).

Although most children are evaluated initially by their primary care physician, the clinician’s diagnostic accuracy for acute sinusitis has not been studied in a primary care setting.

The present study evaluates the usefulness of signs and symptoms in children for the radiological diagnosis of acute sinusitis.

METHODS

The present study was conducted in a university-affiliated hospital over a period of 10 weeks from April to June 1996. All patients (392) were younger than 18 years of age and were consecutive patients for whom a sinus x-ray was ordered.

Patients with chronic sinusitis or symptoms lasting for more than 30 days were excluded from the analysis. The study was approved by the Research Ethics Committee of the authors’ hospital.

Data collection

In the authors’ hospital, a computer-based system is used for ordering all imaging consultations. The request screen was tailored to fit the needs of the study and included a mandatory questionnaire about the probability of the presence of sinusitis, as evaluated by the physician. This probability has to be expressed on a scale of 0% to 100%. Afterwards, the overall clinical probability was classified as high probability (greater than 80%), intermediate probability (40% to 80%) or low probability (less than 40%). In addition, the ordering clinician was required to answer specific questions on the presence of clinical findings. The questionnaire was prepared by a group of four experts (two paediatricians, one family physician and one paediatric allergologist). The authors reviewed the charts for risk factors, physical examination findings and past history of chronic or acute sinusitis. Abnormal nasal examination was defined as the presence of purulent nasal discharge and/or unilateral discharge.

Radiological diagnosis

A sole Waters view was used for children aged younger than 12 years of age (n = 325), and the complete series of four views for older children (n = 62). The remaining five children had an incomplete series for clinical reasons. Radiological characteristics and diagnostic interpretation were provided by two radiologists who were blinded to the clinical data.

On imaging, sinusitis was defined radiographically as complete sinus opacity, an air-fluid level or mucous membrane thickening of at least 4 mm for the maxillary sinus (4). By using the Waters view, mucosal thickening of the maxillary sinuses was measured as the maximum distance from the air-mucosa interface perpendicular to the lateral wall of the maxillary sinus. Mucous membrane thickening was measured at a right angle to the lateral border of a maxillary sinus. The transverse diameter was also measured. By using this classification scheme, high interobserver agreement has been shown (5).

Statistical analysis

Patient demographic characteristics were reported as means ± SD. Clinical data were analyzed initially using the univariate technique. For dichotomous variables, the sensitivity, specificity, positive predictive value, negative predictive value and likelihood ratio were calculated by comparing results with the criterion standard, sinus radiographs (6). Continuous variables (for example, age and duration of symptoms) were analyzed by using the Mann-Whitney U test. Multilevel variables were analyzed by determining conditional likelihood ratios (7). Clinical findings with univariate associations (P<0.25) were analyzed by backward stepwise logistic regression to identify independent predictors of sinusitis (8). Data analyses were performed using Statview 5.0 (SAS Institute, USA).

RESULTS

All of the 392 consecutive patients were enrolled during the study period. Mean age was 5.2 years (SD 4.2 years) (median age was four years). Patients reported a mean symptom duration of 9.5 days (SD 7.6 days).

Two hundred fifty-seven patients (66%) had a radiological diagnosis of acute sinusitis. The severity of radiographic changes is shown in Table 1. One hundred twenty-eight patients (33%) had a chest radiograph. Of those patients, 16 had findings that were abnormal, showing localized opacity consistant with that of pneumonia. Ten patients (2%) had abnormal sinus and chest radiographs.

TABLE 1:

Severity of changes seen on sinus radiographs (n = 392)

Radiographic change	Number of patients (%)
Normal	135 (34)
No abnormality	74 (19)
Musocal thickening less than 4 mm	61 (15)
Abnormal
Mucosal thickening 4 mm or more	257 (66)
Complete opacity	128 (33)
Air-fluid level	14 (4)

The sensitivities, specificities, positive predictive values, negative predictive values and likelihood ratios for clinical evaluation findings are shown in Table 2. No single finding was both sensitive and specific. Unilateral rhinorrhea was highly specific, but few of the patients examined (5%) had unilateral rhinorrhea. Classical symptoms (rhinorrhea lasting more than 10 days and purulent rhinorrhea) increased the likelihood ratio the most (1.3 and 1.34, respectively). High initial probability (greater than 80%), evaluated by the clinician before the radiographic examination, had a likelihood ratio of 2.0 (Table 3). Associated findings for sinusitis most frequently found in patients records were a past history of sinusitis (33%) and asthma (26%) (Table 4). Physical examination findings are shown in Table 5.

TABLE 2:

Performance characteristics of signs and symptoms for the radiological diagnosis of acute sinusitis (n = 392)

Symptoms	Positive LR (finding present)	Negative LR (finding absent)	Sensitivity (%)	Specificity (%)	Frequency (%)	Positive predictive value (%)	Negative predictive value (%)
Rhinorrhea for more than 10 days	1.3*	0.8*	52	60	46	71	40
Productive cough	0.96	1.12	72	25	71	65	30
Interorbitary fullness	1.24	0.85	47	62	32	68	40
Purulent rhinorrhea	1.34*	0.62*	71	47	62	72	46
Unilateral rhinorrhea	1.2	0.99	6	95	5	70	49
Fever above 38.5 ° C	1.0	1.0	46	64	40	70	39

^*95% CI excludes 1 so that P<0.05. LR Likelihood ratio (likelihood ratios were calculated by comparing results with the criterion standard, sinus radiographs)

TABLE 3:

Characteristics of the overall clinical impression

Clinical impression	Sinusitis		Likelihood ratio (95% CI)
	Present	Absent
High probability	31	11	2.0 (1.2 to 3.5)
Intermediate probability	203	96	1.1 (1.0 to 1.2)
Low probability	23	28	0.6 (0.4 to 0.9)

TABLE 4:

Associated findings for acute sinusitis (n = 392)

Associated findings	Number of patients (%)
Past history of sinusitis	129 (33)
Asthma	102 (26)
Allergic rhinitis	15 (4)
Immunodeficiency	7 (2)
Past history of chronic sinusitis	4 (1)
Cystic fibrosis	1 (0.3)
Nasal septum deviation or malformation	1 (0.3)

TABLE 5:

Physical examination findings as reviewed in the patient charts (n = 392)

Physical signs	Present	Absent	Not reported
Fever	103	241	48
Abnormal nasal examination	98	28	266
Acute otitis media	35	357	0
Abnormal sinus palpation	16	37	339
Postnatal drip	14	56	322
Serous otitis	2	390	0
Transillumination	0	0	392

Mucous membrane thickening was correlated with the initial clinical probability of acute sinusitis. The correlation coefficient between mucous membrane thickening and the initial clinical probabilities of acute sinusitis is low (0.17) but statistically significant (95% CI 0.04 to 0.29). Interestingly, initial probabilities higher than 50% are associated with a mean mucous membrane thickening of 4 mm and higher – the usual diagnostic criteria for radiological diagnosis of sinusitis.

Logistic regression analysis showed two independent predictors of acute sinusitis: purulent rhinorrhea (odds ratio 2.0) and the presence of acute otitis media (odds ratio 2.6).

Associations were found between longer duration of symptoms (10.0 compared with 8.6 days, P<0.05) and younger age (4.5 compared with 6.5 years, P<0.05), and a radiological diagnosis of sinusitis.

DISCUSSION

The presence of classical symptoms (rhinorrhea lasting more than 10 days and purulent rhinorrhea) is predictive of acute sinusitis, as diagnosed on sinus radiographs. The physician’s overall clinical impression, expressed as initial probability, was superior to any single historical or examination finding in the diagnosis of acute sinusitis. This finding corroborates a study involving adult patients (2).

Past studies involving children are limited to sensitivities for a few clinical findings. Clear or purulent discharge (sensitivity 76% to 84%) and cough (sensitivity 48% to 80%) were the most sensitive findings but their discriminating power was not known (3,9–11).

The present study used a prospective design in the collection of initial clinical probabilities and presenting symptoms. All children for whom a sinus radiograph was ordered were included. The radiologists were blind to clinical data and used explicit criteria for diagnosing sinusitis (3–4,12). Also, the patients had a broad spectrum of symptoms and were drawn from speciality clinics (ear nose and throat) and general clinics (paediatric, emergency room). The authors’ selection criteria avoided the bias of other paediatric studies in which the criterion standard was applied only to patients with a high probability of disease (3,13).

A potential limitation of the present study is the criterion standard. Sinus aspiration with bacteriologic culture is the gold standard for diagnosing sinusitis, but could not be applied for practical and ethical reasons. Therefore, sinus radiographs and a previously developed classification scheme were used as the criterion standard. This scheme has been shown to correlate well with maxillary sinus aspiration (3,14). Furthermore, sinus radiographs are readily obtained by the primary care physician and can be considered to be a pragmatic standard. Although sinus computed tomography and magnetic resonance imaging are increasingly being used to diagnose sinus disorders and are considered to be more sensitive (particularly for the ethmoid sinuses), they have not been compared with sinus aspiration and may lack specificity (15).

A recent review concluded that no imaging studies are recommended for the routine diagnosis of uncomplicated sinusitis presented to the primary care physician (16). During the present study, 34% of the children did not receive antibiotic agents because they had a normal sinus radiograph. In an era of increasing bacterial antibiotic resistance, selecting patients who really require antibiotic treatment is important.

The prevalence rate of acute sinusitis is high and this must be considered before generalizing results to other institutions where rates may be different.

CONCLUSION

The present study showed that clinical initial impression, expressed in terms of probability, is very useful for diagnosing acute sinusitis. Further studies are required to show whether sinusitis can be ruled out with confidence on the basis of clinical findings. This will be a challenge because it is known that sinusitis in children can be insidious and that there is an overlap of symptoms between sinusitis and the common cold in the paediatric population.