Delay to diagnosis in childhood bone and joint infection

Prompt diagnosis and treatment for childhood bone and joint infection (BJI) is necessary to avoid long-term disease consequences.¹ Chronic infections often require multiple surgeries and can have devastating outcomes.² The exact time frame for irreversible damage is unknown, although in vitro studies report articular cartilage destruction within as little as 8 hours of bacterial inoculation.³

The clinical consequences of chronic and subacute disease are well characterised in the literature, but less attention has been given to the delayed management of acute disease.³ Recent studies focus on diagnostic success in the emergency care setting.^4,5 Epidemiologic studies on BJI in the New Zealand population only report tertiary management.^6,7 There is a lack of information available about prehospital disease characteristics and prehospital medical care. These aspects of childhood BJI are important, as duration of symptoms can predict disease complexity and treatment success.^8–10

Scoring systems to recognise complicated illness incorporate features of hospital presentation, but none include consideration of prehospital delays that contribute to more severe disease.^11,12 For instance, the A-score by Alhinai et al. defines “delayed source control” as 3 days between admission and surgery.¹³ However, it is arguable that a child unwell in the community for a week has also experienced “delayed source control”. The full impact of symptom duration is not measurable unless we account for prehospital disease course.

In addition to this, acute haematogenous osteomyelitis (AHO) and septic arthritis (SA) can be exceptionally difficult to diagnose.¹⁴ For many children, the diagnostic process leads to apparently unavoidable treatment delay. Diagnostic algorithms developed in the 1990s and early 2000s have not been successfully validated in the current clinical environment.⁹ Traditional teaching emphasises fever and limp as indicative of BJI, despite evolving disease aetiology that has altered clinical presentation.¹⁵ In a recent case series only 15% of BJI infections secondary to Kingella kingae (K. Kingae) presented with fever.^16,17 Overlapping clinical presentation with transient synovitis leads to misdiagnosis rates of up to 60% in the emergency care setting.¹⁴ These misconceptions surrounding disease presentation require review and refinement. If they are not revised, misdiagnosis will result in delayed treatment and management for some children, with the consequences being more severe acute disease and increased rates of subsequent disability.

There is likely a temporal relationship between contiguous local infection and duration of symptoms. This relationship has not yet been defined in the literature. Greater utilisation of magnetic resonance imaging (MRI) over recent decades has increased clinical awareness of these contiguous infections.¹⁸ We do not know the exact time frame in which “isolated” osteomyelitis disseminates locally into the surrounding tissues, creating focal collections such as subperiosteal abscess or pyomyositis, which may require more surgery. Children with contiguous local infection experience greater morbidity and incur more direct medical cost.⁸ Determining the temporal nature of disease progression will help us understand the impact of delayed diagnosis.

In summary, the aim of this research is to determine the proportion of delayed diagnosis in childhood BJI and the impact of pre-diagnosis symptom duration on disease trajectory, with focus on length of stay (LOS) and hospitalisation cost.

Methods

This study is a retrospective analysis of all cases of suspected osteomyelitis and SA managed in multiple institutions across Auckland region from 2018–2023. Children from birth to age <16 years were included. Health and Disability Ethics Committee (HDEC) approval was obtained for this study together with institutional review board approval (reference: 19/NTA/46 & 18NTA219).

Comprehensive review of electronic clinical records was conducted. Data were collected on patient demographics, illness presentation and disease type. Household socio-economic deprivation was measured using the New Zealand Index of Deprivation (NZDep), based on patient address at time of treatment.¹⁹ NZDep is an area-based surrogate measure of socio-economic hardship. AHO was defined based on radiographic investigation via MRI or computed tomography (CT) and/or positive intraoperative culture or bone biopsy. SA was defined based upon intraoperative culture results, culture results from aspirate or positive radiographic investigation in the setting of a positive blood culture. Cases of chronic BJI, post-viral or reactive arthritis, post-operative infection, cases associated with significant malignancy or patients with insufficient clinical data for analysis were excluded.

Children were considered to have multifocal sepsis if they had multiple concurrent infections; for example, AHO with pneumonia. Locally contiguous infection was defined as AHO with contiguous pyomyositis, subperiosteal abscess or adjacent joint SA.

Electronic hospital data contains primary and urgent care referrals, community laboratory results and community pharmacy dispensing records. These were reviewed to describe medical attention prior to hospital admission. In New Zealand, all medications dispensed to children from community pharmacies are listed electronically.²⁰

Cases were assessed for any alternative diagnosis prior to identification of BJI. For example, if they presented to primary care or the emergency department (ED) and were discharged with a diagnosis of “transient synovitis” but subsequently confirmed to have AHO or SA. For the purposes of this study, diagnostic delay is defined as those children with symptoms for >3 days before treatment initiation.

LOS, number of surgeries and duration of combined intravenous and oral antibiotic therapy were recorded. The cost of hospitalisations was determined using a weighted discharge value (i.e., Weighted Inlier Equivalent Separations [WIES]) for all National Minimum Dataset events as calculated by the New Zealand Ministry of Health – Manatū Hauora.²¹ This paediatric WIES cost-weight encompasses medical costs, ward stays, medications, laboratory investigations, operations and nursing and other ward staff. It is based on diagnostic-related groups (DRGs), with additional costs for interventions such as mechanical ventilation.

Results

An initial 994 encounters for BJI were identified by clinical codes for pyogenic arthritis and osteomyelitis. Of these, 563 met criteria for first presentation with acute BJI. The majority of exclusions were readmissions for the same condition or chronic disease (n=192), incorrect diagnosis (n=74), or infections secondary to operative complication or penetrating injury (n=56). Eligible cases were primarily AHO (39%); multifocal sepsis was less frequent (16%) (Table 1).

A presenting complaint was clearly outlined in 512 cases. A total of 218 children (43%) received an alternative diagnosis before identification of BJI (Table 2). A “classical” presentation with fever and pain at the site of infection was only seen in 45%. History of a recent viral illness was common (39%), as was history of trauma (23%).

An alternative diagnosis was less common in children of Māori or Pacific ethnic groups (Table 2). There was a lower average NZDep in children who received another diagnosis first (NZDep 6 vs 7, p=0.0003). Of note, within our cohort, the proportion of NZ European patients attending community medical centres was slightly higher (n=116, 66%) than Māori (n=40, 48%) or Pacific (n=69, 45%) (p=<0.01).

Children who were alternatively diagnosed were more likely to have presented to a primary or urgent care centre (82% vs 38%, p=<0.0001). They were equally likely to have reported fever and pain at infection site (47% vs 44%, p=0.62). However, they were more likely to have a recent viral illness (46% vs 34%, p=0.008).

Delay between symptom onset and treatment for BJI was greater if children received another diagnosis first (7.8 vs 4 days, p=<0.0001). Once hospitalised, children with initial misdiagnosis were equally likely to require surgery (50% vs 53%, p=0.56). Pathogen type in those with an alternative first diagnosis was more likely culture-negative or K. kingae–mediated BJI (58% vs 47%, p=0.004).

Considering diagnostic delay, Figure 1 shows the duration of symptoms prehospital for children with and without an alternative first diagnosis. Those without alternative first diagnoses had shorter duration of symptoms before treatment. Diagnostic delay of >3 days was more common in those with an alternative first diagnosis (75% vs 33%, p=<0.001).

The relationship between duration of symptoms and disease subtype is demonstrated in Figure 2. This Kaplan–Meier survival analysis compares days of symptoms before diagnosis across different subtypes of disease. It shows high proportion of SA and multifocal sepsis diagnosed after a short duration of symptoms. AHO and locally contiguous infections remain undiagnosed for longer.

Examining for statistical significance, children with multifocal sepsis typically presented within 48 hours, or “early” (Table 3). Conversely, locally contiguous infections were more often diagnosed “late” (>1 week of symptoms) (p=0.02). For the entire cohort, surgical intervention rates appear higher for those who present during the first 48 hours of symptoms (50–60%) or after 10 days of symptoms (64%) (p=0.09) (Figure 3).

Surgical intervention, LOS and cost differed by disease subtype. In Table 4, these outcomes for children with “isolated” AHO have been compared to locally contiguous infections. This is for two reasons. Firstly, it is already established that patients with SA and multifocal sepsis generally are more likely to require surgery, which raises hospitalisation cost.²² Operative drainage of SA is deemed gold standard in the majority of cases.^23,24 Secondly, it is possible that earlier recognition and treatment of AHO could prevent progression to locally contiguous infection. Any additional costs and morbidity associated with contiguous infection may be avoidable.

Our results showed the surgical intervention rate for contiguous disease was higher than for isolated AHO (70% vs 14%, p=0.0001) (Table 4). Children with locally contiguous infection spend longer in hospital (8.42 vs 5.14 days, p=<0.0001). Hospitalisation cost appeared higher but did not reach statistical significance (NZ$14,102 vs NZ$8,587, p=0.26). A large standard deviation in cost averages is due to high additional costs associated with paediatric intensive care admission.

Discussion

Overall, in this study almost half of children with BJI were given an alternative diagnosis first. Alternatively diagnosed children were more likely to have attended primary or urgent care centres and report a recent viral illness. An alternative diagnosis delayed treatment by an average of 3.8 days.

Greater duration of symptoms is associated with locally contiguous forms of BJI. The consequences of this can be seen clinically as increased need for surgical intervention and longer hospital stay. Hospitalisation for locally contiguous infections was on average 3 days longer, with direct hospitalisation costs almost twice as high as for “isolated” AHO. Although this difference in hospitalisation cost did not reach statistical significance, it is likely clinically meaningful; longer hospitalisation also confers additional treatment expense to patients and families.

Delay in making a diagnosis may reflect outdated understanding of disease aetiology. Transient synovitis, unlike bacterial infection, is commonly associated with a pre-existing respiratory illness.^14,25 Traditional teaching emphasises the relationship between viral illnesses and transient synovitis, without acknowledging a potential relationship between viral infections and bacterial BJI.²⁶ A viral infection may be concomitant with Staphylococcus aureus (S. aureus) or S. pyogenes (Streptococcus pyogenes) mediated AHO or SA, whereby initial viral infection lowers immunity to allow bacterial invasion.²⁶ This is supported by the lower rates of bacterial bone infections during the COVID-19 pandemic, as well as the high incidence of pre-existing viral infection in our cohort.²⁷ For clinicians, recognising the association between viral infections and childhood BJI may improve diagnostic rate. A recent virus may in fact be associated with BJI, rather than pointing to a diagnosis of transient synovitis.

This study suggests less clinical awareness for disease presentation associated with low virulence organisms and, in particular, culture-negative and K. kingae–mediated infections. Our results confirmed the proportion of culture-negative and K. kingae–mediated disease is higher in our region compared to previous years.^6,28 Emergence of K. kingae has been attributed to widespread vaccination for Haemophilus influenzae since the 1980s.¹⁷ In addition, daycare use has increased, which is associated with spread of K. kingae.^29,30 These children often appear clinically well, increasing reliance on laboratory and radiographic information for diagnosis. Those with culture-negative or K. kingae infections were more likely to experience diagnostic delay in our cohort, perhaps in association with subtle disease presentation.

In contrast, S. aureus–mediated BJI is classically associated with febrile illness, has increased representation in lower socio-economic groups and is over-represented in Māori and Pacific children.³¹ In our study delayed diagnosis was less common for these ethnic groups; although, it is important to note that Māori and Pacific children in our region experience reduced primary healthcare access.³² Māori and Pacific children with BJI are also more likely to have severe illness and receive treatment in the paediatric intensive care unit (PICU).³³ To improve diagnostic success across a range of disease presentations, contemporary teaching for identification and management of BJI needs to reflect changes in clinical epidemiology.

The consequence of diagnostic delay in childhood BJI is delayed treatment initiation, specifically intravenous antibiotics and surgical source control. Importantly, diagnostic delay was equally likely among children who presented with fever and pain at site of infection in this cohort. Furthermore, children who received an alternative diagnosis first went on to receive equivalent rates of surgical intervention, suggesting delays were not limited to mild forms of disease. Although management guidelines vary between institutions, evidence supports urgent antibiotic administration in cases of musculoskeletal sepsis.³⁴ If osteomyelitis or SA are part of the disease differential, timely referral to a tertiary centre with radiographic and laboratory analysis may support earlier treatment initiation.

A limitation of this research is the quality of documentation. In electronic case records, a recent viral illness or trauma could be deemed irrelevant by clinicians and not documented. As a consequence, prevalence of these may be under-reported in this retrospective case series. However, we have sought to reduce the potential for this bias by careful analysis of primary care referral documentation and community laboratory and dispensing records where available. It is possible that this does not capture all consultation data, particularly where a prescription is not filled.

There are several findings from this study that can inform ongoing clinical practice. Firstly, models of teaching and diagnosis for BJI should reflect current clinical and molecular epidemiology of disease. This means that ongoing collection of data around presentation and causative pathogens is essential to inform surgeons and the wider medical community. Knowledge of how disease presents in the current environment is crucial to minimise delays to diagnosis and misdiagnosis.

Secondly, without accessible imaging and laboratory testing it is very difficult to diagnose childhood BJI. This is reflected in the greater delay to treatment and diagnosis after presentation to primary care. With any clinical suspicion of AHO or SA, referral to a facility with imaging and laboratory capability is essential. Healthcare professionals need to understand that the goal is to prevent contiguous infection, which develops over time and is associated with greater surgical intervention and morbidity. For patients and families, the accumulation of direct medical costs and productivity loss is greater with delayed diagnosis.

Conclusion

Childhood AHO and SA are frequently given an alternative diagnosis at presentation, leading to delayed treatment initiation. Delays of >1 week are associated with contiguous infection, which more frequently requires surgery, with prolonged morbidity and cost.

View Figure 1–3, Table 1–4.