Abstract
Background Severity assessments of respiratory tract infection (RTI) in children are known to differ between parents and clinicians, but determinants of perceived severity are unknown.
Aim To investigate the (dis)agreement between, and compare the determinants of, parent and clinician severity scores.
Design and setting Secondary analysis of data from a prospective cohort study of 8394 children presenting to primary care with acute (≤28 days) cough and RTI.
Method Data on sociodemographic factors, parent-reported symptoms, clinician-reported findings, and severity assessments were used. Kappa (κ)-statistics were used to investigate (dis) agreement, whereas multivariable logistic regression was used to identify the factors associated with illness severity.
Results Parents reported higher illness severity (mean 5.2 [standard deviation (SD) 1.8], median 5 [interquartile range (IQR) 4–7]), than clinicians (mean 3.1 [SD 1.7], median 3 [IQR 2–4], P<0.0001). There was low positive correlation between these scores (+0.43) and poor inter-rater agreement between parents and clinicians (κ 0.049). The number of clinical signs was highly correlated with clinician scores (+0.71). Parent-reported symptoms (in the previous 24 hours) that were independently associated with higher illness severity scores, in order of importance, were: severe fever, severe cough, rapid breathing, severe reduced eating, moderate-to-severe reduced fluid intake, severe disturbed sleep, and change in cry. Three of these symptoms (severe fever, rapid breathing, and change in cry) along with inter/subcostal recession, crackles/crepitations, nasal flaring, wheeze, and drowsiness/irritability were associated with higher clinician scores.
Conclusion Clinicians and parents use different factors and make different judgements about the severity of children’s RTI. Improved understanding of the factors that concern parents could improve parent–clinician communication and consultation outcomes.
INTRODUCTION
Respiratory tract infection (RTI) is the most common reason parents take children to consult in primary care worldwide,1 and many of these children are prescribed antibiotics unnecessarily. The decision to seek medical advice is complex,2,3 as parents rely on their child’s physical expression of illness and their own interpretation of the symptoms.4 Clinicians also report consultations for RTI as being complex because they manage clinical uncertainty regarding diagnosis and prognosis, which often results in ‘just-in-case’ antibiotic prescribing.5
Patient-centred care has been shown to improve the quality of doctor–patient communication, consultations, and illness outcomes,6 but reaching a shared understanding of the nature of the problem and its severity is central to such care.7,8 Illness severity can best be described as the magnitude of the patient-perceived, clinically significant manifestations of disease processes that are associated with decrements in health-related quality of life or health status.9 In children’s consultations, there are three interacting perceptions — that of the child, parent, and clinician — which affect the assessment of illness severity, and assessments are likely to vary due to the different factors taken into account.
Parent-perceived severe illness in children with RTI is one reason why parents choose to consult in primary care,10 and a clinician’s illness severity assessment might be the guiding factor for antibiotic treatment choice. Misinterpretation of a child’s illness severity might not only lead to overconsultation, but also to overtreatment. In-depth knowledge of the differences in the factors that determine illness severity assessment — and, particularly, (dis) agreement between parents’ and clinicians’ assessments — may be important in improving parent–clinician communication and the management of children with RTI in primary care. As such, this study aimed to:
investigate (dis)agreement between clinicians’ and parents’ illness severity scores; and
identify and compare the determinants of high parent and clinician severity scores for children presenting to primary care with cough and RTI.
How this fits in
The decisions parents make regarding the seeking of medical advice are complex. An improved mutual understanding of the factors taken into account when assessing illness severity could facilitate better parent–clinician communication and improved parental understanding about the symptoms that necessitate consultation.
METHOD
Design and study population
Data were used from the TARGET study,11 a multicentre prospective cohort study of children with acute (≤28 days) cough and RTI recruited between July 2011 and May 2013. The design of that study has been described in detail and the main results published elsewhere.11,12 In brief, GPs and prescribing nurse practitioners (hereafter referred to as ‘clinicians’) working in primary care centres — GP surgeries, walk-in centres, GP out-of-hours centres, or polyclinics — were recruited and trained by four UK hubs (Bristol, London, Oxford, and Southampton). The clinicians recruited children to the study if they were eligible, which was defined as:
Children with an infected exacerbation of asthma and those who were severely unwell (for example, those who required same-day hospital assessment or admission) were included. Children were excluded if they:
presented with a non-infective exacerbation of asthma;
were at high risk of serious infection (immunocompromised, for example, with cystic fibrosis);
required a throat swab for clinical management (which was taken for research purposes in a subgroup of children);
had been previously recruited to the study;
had recently participated in other research; or
had registered at the practice temporarily.
Measurements
After obtaining informed written parent consent, clinicians completed a structured online or paper-based case report form (CRF) (available from the authors on request). The form recorded:
seven sociodemographic items;
four illness history/trajectory items;
29 parent-reported symptoms (including whether mild, moderate, or severe in the previous 24 hours);
14 physical examination signs (including vital signs); and
the prescription of antibiotics (none, immediate, or delayed) or referral to secondary care for acute assessment.
Children’s diagnosis of current asthma was checked by reviewing medical notes; asthma was deemed to be present if it had been recorded in the notes and asthma medication had been issued in the previous 12 months. Socioeconomic status, based on the patient’s postcode, was assessed using the Indices of Multiple Deprivation (IMD), with scores ranging from 0 (least deprived) to >90 (most deprived). Illness severity was measured independently by parents and by a clinician using a 0–10 visual analogue scale (VAS), as outlined by McCormack et al,13 ranging from 0 (well) to 10 (very unwell).
Variables
Illness severity scores
In the main analysis, children were coded as having ‘high parent-reported illness severity’ if the severity score, as assessed by their parent, was ≥7; ‘high clinician-reported illness severity’ was coded if the clinician’s score totalled ≥4. For subanalysis, three illness severity groups were used to determine agreement between the parent’s and the clinician’s scores:
parent score < clinician score: ‘less ill’;
parent score = clinician score: ‘as ill’; and
parent score > clinician score: ‘more ill’.
Symptom severity
For symptom severity (mild, moderate, or severe) in the 24 hours prior to consultation, each variable was split depending on the prevalence:
Clinical cut-offs
In case of dichotomising, commonly used clinical cut-offs were used for continuous data where possible (for example, temperature ≥37.8°C). Age-related heart rate, respiratory rate, and oxygen saturations were coded as raised or normal/low according to Advanced Paediatric Life Support: the Practical Approach.14
Data analysis
The clinician score data were heavily skewed so a non-parametric approach that makes no assumption about distribution was chosen. Illness severity scores from parents and clinicians were compared using the non-parametric approach for unpaired data. Medians, interquartile ranges (IQRs), and the Mann–Whitney test were used to investigate differences between group scores (lower/equal/higher).
To determine the association between the number of symptoms and signs reported in the global illness severity scores, correlation was determined between the parent illness severity score and the number of parent-reported symptoms, and between the clinician illness severity score and the number of symptoms and signs (both separately and added together). Correlation and agreement were calculated using Spearman’s rho (r) and kappa (κ) statistics respectively. To control for anchoring, the rounded mean difference between the parent’s score and the clinician’s score was added to the clinician’s score and κ recalculated.
Univariable and multivariable logistic regression was used to identify the sociodemographic and clinical factors independently associated with parent- and clinician-reported illness severity. Multivariable models were derived through several iterations using backward stepwise logistic regression, including all variables that were statistically significant in the univariable analyses (P<0.01) where missing data were <1%. The authors controlled for age, sex, and ethnicity, as well as the presence of other symptoms or signs already in the model.
Two separate multivariable models were determined for high clinician score: one included demographics and symptoms (for comparison with the parental model), and the other included demographics, symptoms, and clinical signs. The latter clinician model included a combined variable for parent-reported severe fever or clinician-measured temperature ≥37.8°C, and a combined variable for parent-reported rapid breathing or clinician-measured raised respiratory rate. The final models were constructed in the following order:
To reduce the problem of multiple comparisons in the analysis, each listed symptom in the CRF was only tested in terms of whether it was present during the illness, and had been moderate or severe in the previous 24 hours.
RESULTS
In total, 8394 children were recruited to the study across the four centres. There was no difference in parent-reported illness severity scores between the 164 children declining participation (median 3, IQR 2–4) and the final recruited sample (median 3, IQR 2–4). For those declining participation, global illness severity scores were accrued as a result of a clinician’s recruitment log, where clinicians were asked to record reasons for not inviting potentially eligible children to participate and to report the same global illness severity scores.
Children’s median age was 3 years (IQR 1–6 years) (Table 1), with 1392 (16.6%) aged <1 year. In total, 51.6% were boys and 78.4% were white; the mother’s median age at the child’s birth was 30 years, and 18.0% of recruited children’s mothers were current smokers. Families’ median deprivation score was 16.7 (IQR 8.8–29.5). Ethnicity, deprivation score, and prevalence of maternal smoking were similar to national figures.15,16 The median illness duration at recruitment was 5 days (IQR 3–10 days, range 0–28 days); the median number of parent-reported symptoms was seven (IQR 5–9, range 1–16) prior to consultation and six (IQR 4–8, range 0–16) within 24 hours of the consultation (Table 1).
Table 1. Characteristics of children (n = 8394) and parents
Illness severity
All in all, 8360 (99.6%) children had complete parent and clinician illness severity score data. Parent severity scores (mean 5.2 [standard deviation {SD} 1.8], median 5, range 0–10, IQR 4–7) were higher (P<0.0001) than clinician scores (mean 3.1 [SD 1.7], median 3, range 0–9, IQR 2–4) (Figure 1, Table 2). Parents scored illness severity lower than clinicians in 5.7% of cases, the same in 14.8% of cases, and higher in 79.4% of cases (Table 2).
Figure 1. Distribution of illness severity scores.
Table 2. Parent and clinician illness severity scores
The authors found evidence of a low positive correlation (Spearman’s r 0.43, P<0.001) and poor inter-rater agreement between the parent and clinician illness severity scores (κ 0.049) (Figure 2), the parent scores and the number of parent-reported symptoms (Spearman’s r 0.37), and the clinician scores and the number of parent-reported symptoms (Spearman’s r 0.34). The clinician scores and the number of clinical signs are strongly positive correlated (Spearman’s r 0.71); moderate positive correlation was found between clinician scores and the number of parent-reported symptoms plus the number of clinical signs (Spearman’s r 0.60). Anchor controlling for the inter-rater agreement did not improve κ (0.064) (Figure 2).
Figure 2. Comparison of clinician and parent illness severity scores.
a To control for anchoring, the rounded mean difference between the parental score and clinician score was added to the clinician score.
Univariable analyses
All but one variable (oxygen saturation) had <1% data missing. There was no evidence of differences between children with high parent illness scores compared with those with low scores regarding age, sex, ethnicity, number of consultations in the previous 12 months, and current asthma. There was no evidence of differences between children with high clinician illness scores compared with those with low scores regarding ethnicity, number of consultations in the previous 12 months, and parent-reported moderate-to-severe diarrhoea (data not shown). Table 3 summarises the sociodemographic and clinical factors that were associated with higher and lower parent and clinician scores.
Table 3. Univariable associations (P<0.01) with parental and clinician illness severity score
Multivariable analyses
Parent severity scores
The model included 8208/8394 (97.8%) of the children in the cohort. Eight predictors were strongly associated with parent severity scores at P<0.001:
one demographic variable, ≥2 children at home (odds ratio [OR] 1.28, 95% CI = 1.15 to 1.44); and
seven parent-reported symptoms, in order of importance:
— severe fever (OR 2.58, 95% CI = 2.12 to 3.13);
— severe dry cough (OR 1.93, 95% CI = 1.60 to 2.34);
— rapid breathing (OR 1.88, 95% CI = 1.69 to 2.10);
— severe reduction in eating (OR 1.58, 95% CI = 1.26 to 1.98);
— moderate-to-severe reduced fluid intake (OR 1.55, 95% CI = 1.34 to 1.80);
— severe disturbed sleep (OR 1.32, 95% CI = 1.14 to 1.52); and
— change in cry (OR 1.30, 95% CI = 1.13 to 1.49).
Together these gave an area under the receiver operating characteristic curve (AUROC) of 0.68 (95% CI = 0.66 to 0.69) (Table 4).
Table 4. Final multivariable predictors of high parent illness severity score and high clinician illness severity score (all P<0.01)
Clinician severity scores
Two multivariable regression models were derived: one to compare with the parental model including demographics and parent-reported symptoms only (model one), and one including demographics, parent-reported symptoms, and clinical signs (model two). Both models included 8198/8394 (97.7%) of the children in the cohort. Model one identified four predictors with P<0.001, the last three of which were reported by a parent:
illness deterioration recently before consultation (OR 2.11, 95% CI = 1.89 to 2.34);
severe fever (OR 3.04, 95% CI = 2.52 to 3.67);
rapid breathing (OR 1.78, 95% CI = 1.61 to 1.98); and
change in cry (OR 1.29, 95% CI = 1.14 to 1.46).
The combined AUROC was 0.67 (95% CI = 0.66 to 0.68) (Table 4).
Adding clinical examination did not change model-one factors included but identified six additional signs. In order of importance, these were:
inter/subcostal recession (OR 4.91, 95% CI = 3.54 to 6.82);
crackles/crepitation (OR 4.79, 95% CI = 4.18 to 5.50);
nasal flaring (OR 3.08, 95% CI = 1.76 to 5.41);
wheeze (OR 2.31, 95% CI = 1.98 to 2.68);
irritability/drowsiness (OR 1.91, 95% CI = 1.22 to 2.99); and
pallor (OR 0.58, 95% CI = 0.49 to 0.70).
Together these gave an AUROC of 0.79 (95% CI = 0.78 to 0.80) (Table 4).
As clinicians were only able to take the oxygen saturation level for just over half of the children recruited (n = 4194, 50%), a multivariable model was derived using this smaller number of children. This sensitivity analysis did not substantially change the final models (data not shown).
DISCUSSION
Summary
Parent and clinician global illness severity assessment differed, with parents considering their child more unwell than clinicians. Factors associated with illness severity also differed between parents and clinicians, with parents relying on symptoms and clinician’s physical examination findings. That said, severe fever and rapid breathing were important for both parents and clinicians.
Strengths and limitations
The data used were from a well-characterised, large (England — London, Bristol, Southampton, and Oxford), representative cohort of children presenting to primary care with the most common problem managed by primary care health services. Participating children had similar levels of overall illness severity as those who were invited to participate in the study but declined to do so. Baseline characteristics were pragmatic, measured according to routine clinical practice, and had high level of completeness. The main outcome is clinically relevant and a stringent model retention criterion (P<0.01) was used because of the many candidate predictors. The study question described in this article was not the focus of the cohort study but a secondary hypothesis.
The main limitation was the use of a VAS to measure illness severity. This provides a simple technique for measuring subjective experience and has been established as valid and reliable in a range of clinical and research applications;13 however, although ease of use is frequently cited as a major advantage over other scales, it has been pointed out that a VAS requires the ability to transform a complex assessment into a visual-spatial display, involving perceptual judgement and accuracy.17 The VAS used in this study was anchored by 0 (well) and 10 (very unwell). During decision making, anchoring occurs when individuals use an initial piece of information to make subsequent judgements. Once an anchor is set, other judgements are made by adjusting away from that anchor. Anchor points of parents and clinicians might differ as clinicians are likely to be used to seeing many more ill children than parents (none of the clinicians gave a score of 10/10 [very unwell]), which might lead to a lower score of illness severity in general. As such, the poor agreement found between parent and clinician scores might be partly due to anchoring. However, agreement remained poor when adjustments were made for changes in anchoring and the authors found that different parent and clinician factors were associated with high severity scores.
Information on any prior assessment (for example, GP triage, NHS 111 triage, or pharmacist recommendation) that might have influenced a parent’s decision to consult a GP was not included in the analyses but could have consequently influenced the parent’s illness severity assessment.
Another limitation is that, although parents are taught to look out for signs of respiratory distress by GPs and 111 in general, comparisons were not like with like: parents did not clinically examine their child and might not be aware of clinical signs such as inter/subcostal recession. As such, modelling the drivers of severity scores was always going to give different results. However, this does reflect the actual situation that parents and clinicians must navigate.
Comparison with existing literature
The poor agreement between parent and clinician illness severity assessments found in the study presented here is consistent with the results of a previous systematic review of qualitative evidence on the interaction of primary care consultations for children with acute minor illnesses.18 That review concluded that common misunderstandings occur when parents and clinicians speak about the ‘seriousness’ of the illness, with parents and clinicians talking at cross-purposes: parents are seeking to justify their decision to consult, whereas clinicians seek to justify non-antibiotic treatment strategies.18
Although the analysis presented here showed that clinicians do take into consideration parent-reported symptoms, they seem to rely on clinical examination. The exploratory analysis on the correlation between the number of symptoms and clinical signs reported and the illness severity scores showed that the number of parent-reported symptoms does not influence the parent score, but the number of clinical signs does influence the clinician score. As such, the authors believe that the type of parent-reported symptom may be more important than the number, but for clinicians it could be a case of ‘the more clinical signs, the more they worry’ (meaning a higher score). This is consistent with qualitative evidence showing that GPs not only rely on the initial assessment, but also feel the need for a more deductive assessment, including physical examination, to refine their diagnosis and rule out serious illness.5 A study by Blacklock et al, in which symptoms and clinical features correctly identify serious respiratory infection in children attending a paediatric assessment unit, showed that parent-reported symptoms were unreliable discriminators of serious acute respiratory infection in children.19 Nurse trial assessment of respiratory distress and some vital signs, such as a temperature of ≥40°C, have also been found to be important predictors, which is in line with the findings presented here.19
The main analyses from this cohort study found that parent-reported fever and clinician-measured inter/subcostal recession, as well as wheeze on auscultation, are prognostically significant; these factors have been shown to have predictive utility in identifying children at risk of future hospitalisation resulting from their RTI.12 These factors are also included as intermediate/red flags in the traffic-light system for identifying risk of serious illness published by the National Institute for Health and Care Excellence (NICE):
It is possible that parents have intuitively identified these factors as important and, therefore, also use them to assess illness severity.
Nothing was found in the literature to explain the counterintuitive findings of pallor and one child in the home being associated with lower illness severity scores. Contrary to the finding presented here, pallor is one of the red-flag symptoms identified in the NICE traffic-light system for identifying risk of serious illness in children with feverish illness.
Implications for practice
Interventions to promote self-care — one of the cornerstones of the NHS’s sustainability strategy — should take account of the concerns that are likely to be the main drivers of parents seeking medical help. Some symptoms that are of importance to parents (severe dry cough, reduced eating, and disturbed sleep) have not shown to be predictive of poor outcome12 and may be important targets for reassurance, for example, the likelihood of a parent perceiving illness to be severe is 14–52% higher for parents if their child’s sleep has been disturbed but disturbed sleep was not a predictor of illness severity for clinicians. Moreover, how to identify and address these concerns should be included in primary care clinician (medical and nursing) training programmes, as recommended by Health Education England’s recent strategy document, Tackling Antimicrobial Resistance.21
The study presented here showed that fever is associated with both high parent and clinician illness severity scores. Education on managing fever for parents and clinicians is key to making better-informed decisions as to when, or if, to consult and improving parent–clinician communication. In some cases, parent concerns regarding fever are the result of lack of experience and knowledge;22,23 empowering parents and teaching them about alarm symptoms could ameliorate illness anxiety and, possibly, improve use of primary care services.24 Advice on self-management for parents and carers could be provided to help patients self-manage fever.25 However, severe fever in a primary care setting is also associated with serious infection26 so parents may be correct to be concerned; as such, fever and fever management should be an important item in guidelines (produced, for example, by NICE) and telephone triage protocols (for example, via NHS 111) on acute RTI.
Clinicians need to recognise that parents reach different conclusions and use different factors when making judgements about illness severity. Improved understanding of the factors that concern parents could improve parent–clinician communication and the quality of consultations, thereby also improving child health outcomes. Understanding parents’ concerns and educating them about clinicians’ concerns need to be part of day-to-day practice; balancing the two is essential to successfully manage patients with minor infection. Parent education should continue to be an important element of clinical care.
