Paediatric care in general practice: Case mix, referral patterns and healthcare costs

General practitioners (GPs) provide most healthcare to Australian children,¹ referring to public and private specialists. However, wait times to be seen in hospital emergency departments (EDs) and outpatient clinics for paediatric care are increasing.² Similarly, private paediatricians report long wait times or closing their books, especially to children with developmental–behavioural concerns.³ The burden of this care likely falls back on GPs, but we have no contemporary data to describe which paediatric conditions GPs see, which conditions they refer (and where) and whether referrals differ by general practice, patient or GP factors. We also do not understand the cost implications of GP paediatric referrals to the healthcare system.

A better understanding of who GPs care for, who they refer and for what reasons could inform GP training and broader system reform aimed at strengthening the comprehensive healthcare of Australian children close to home. Further, it could help identify any inequity in access to specialist care, informing new models of care to optimise access for all children.

Therefore, the aims of this study, performed in 22 general practices from Victoria and New South Wales (NSW), were to describe: the case mix of children aged between 0 and <18 years attending GPs (Aim 1); patterns of GP referrals for children (Aim 2); practice, GP or patient factors associated with referrals (Aim 3); and costs to the healthcare system (Aim 4).

Methods

This paper draws on baseline data from the Strengthening Care for Children (SC4C) trial (Australia New Zealand Clinical Trials Registry 12620001299998).⁴ Briefly, we recruited within the catchment areas of the North Western Melbourne Primary Health Network in Victoria and the Central Eastern Sydney Primary Health Network in NSW. These primary health networks had high paediatric referral rates to local children’s hospitals and were interested in participating in the study. Participating general practices needed to be accredited by the Royal Australian College of General Practitioners (RACGP) or working towards accreditation, have 900 or more paediatric patients attending the practice in the last 12 months and sign a licence agreement to install the software clinical data extraction tool, GRHANITE™, in their electronic medical record (EMR).⁴ GRHANITE enabled the extraction of de-identified, routinely collected data, including patient demographics (sex, age, socioeconomic status based on Socio-Economic Indexes for Areas [SEIFA] quintile), reason for visit, consultation length and Medicare item billing. GP referrals are not consistently recorded in EMRs; therefore, the GRHANITE team developed a tailored pop-up window for Victorian and NSW locations (Appendix 1, available online only), displaying referral outcomes that GPs completed following each paediatric consultation. This pop-up window featured the most common referral options, including no referral or referral to public or private services. GPs completed a baseline survey measuring demographics, number of years in practice, number of children seen per week, any formal training in paediatrics, factors influencing their decision to refer children and their use of HealthPathways (an online information portal for primary healthcare providers).

For Aim 1, ‘reason for visit’ was categorised based on a previous pilot⁵ by two of the authors who are paediatricians (HH, RL). A natural language processing algorithm was then developed with The University of Melbourne to automatically transform general practice EMR clinical text of ‘reason for visit’ or diagnosis into structured clinical data, based on Systematized Nomenclature of Medicine Clinical Terminology (SNOMED CT).⁶ Where we found abbreviations and misspellings (eg upper respiratory tract infection, URTI, upper resp infection), two paediatricians reviewed these and then determined how they best mapped to available SNOMED CT codes using the CSIRO OntoServer system from the Australian National Clinical Terminology Service.⁷ The following categories were developed: medical; developmental–behavioural; immunisation; mental health; and encounter for check-up. For each category, we calculated the frequency and proportion of clinic visits. Within each category, we then calculated the top 10 conditions encountered, the length of clinic visit (median and interquartile range [IQR]), the most frequent Medicare billing item number and the number and percentage of consultations that resulted in a referral.

For Aim 2, we calculated the number and proportion of consultations where the patient was referred and their referral destination overall, as well as by reason for visit. Referral destinations were collapsed across the two states into public ED, public outpatient, other public hospital, private hospital, private specialists, public mental health services, allied health and other.

For Aim 3, we described the associations between general practice, GP and patient factors and the likelihood of referrals using risk ratios (RRs) and 95% confidence intervals (CIs) calculated using univariate mixed effects Poisson regression with robust error variance. Each model includes a random effect for GP and general practice to allow for clustering in data. GP practice factors included practice billing type for paediatric patients (ie private [no patients offered bulk billing], mixed private and bulk billing and bulk billing only). GP factors included sex, years of practice, number of children seen per week, prior formal paediatric training, use of HealthPathways and factors influencing the decision to refer a paediatric patient. Patient factors included age, sex and socioeconomic status. Statistical analyses were conducted using Stata 16.0.

For Aim 4, we applied a frequency and unit cost to each referral identified in Aim 2. Frequencies of specialist visits were obtained from analysing population Medicare data from the Longitudinal Study of Australian Children for number of initial child specialist visits in a year and the number of review visits in a year.⁸ The mean number of psychologist visits was taken from Hiscock et al,⁸ with allied health visits assumed to be the same. For each visit, a unit cost associated with a closely aligned Medical Benefits Schedule item number⁹ was applied. The number of referrals was multiplied by the mean number of visits and unit cost to produce a total cost. A sensitivity analysis was performed modifying the proportion of children who would take up that referral. Costs are reported in 2023 Australian dollars.

This study was approved by The Royal Children’s Hospital Ethics Committee (August 2020; Project ID: 65955) and site-specific human research ethics committees.

Results

Of the 22 general practices, 15 (68.2%) were bulk billing, four (18.2%) were mixed billing and three (13.6%) were exclusively private billing. A median of 5.5 (minimum 3 to maximum 12) GPs per practice consented to participate. The characteristics of the participating GPs are presented in Table 1. Compared with GPs in non-participating practices in the two primary health networks, more female GPs participated in in our trial (58% vs 51%) and a lower percentage of GPs had been practising for more than 15 years (36% vs 58%). There was no difference in practice socioeconomic status (based on SEIFA of practice postcodes) between practices that did and did not participate.

Aim 1: Case mix of children aged between 0 and <18 years attending general practices

The total number of clinic visits from 1 May 2021 to 31 March 2022, as well as visit characteristics, are presented in Table 2. Of 49,932 consultations, medical issues were the most frequent reason for the visit (n=29,289), followed by immunisations (n=7745), developmental–behavioural (n=1170), encounter for check-up (n=1143) and mental health (n=888). In terms of referrals by reason for visit categories, GPs most commonly referred for mental health problems (33.7% of all mental health consultations) followed by developmental–behavioural concerns (26.8%).

Aim 2: Referral numbers, proportions of consultations and destination

Referral destinations according to the reason for visit category and according to the top five reasons for the visit within each category are presented in Table 3. GPs referred 4420 of 43,301 (10.2%) children. Within the mental health and developmental–behavioural reason categories, most referrals were to private specialists, especially for attention deficit hyperactivity disorder (40.0% of consultations referred to private specialist), behaviour problems (37.1%), anxiety (31.1%) and autism spectrum disorder (26.9%). Referrals to allied health were common for speech delay and behaviour problems. Referrals to hospital clinics or to EDs were uncommon (2.4%).

Aim 3: Factors associated with referrals

Factors associated with GP referrals are shown in Figure 1 (refer also to Appendix 2, available online only). Factors associated with an increased likelihood of GP referral included the practice being private billing (vs bulk billing; risk ratio [RR] 1.72; 95% confidence interval [CI] 1.23–2.39; P=0.001), being a female (vs male) GP (RR 1.30; 95% CI 1.12–1.52; P=0.001), the GP reporting that they were sometimes or frequently (vs rarely) influenced in their decision to refer by the belief that a procedure could only be provided by a paediatrician (RR 1.67 [95% CI 1.18–2.38; P=0.004] and RR 1.54 [95% CI 1.09–2.18; P=0.014, respectively) and the child being older than other children seen by GPs and male.

Factors associated with a reduced likelihood of GP referral include the GP practising for 6–15 (vs <6) years (RR 0.77; 95% CI 0.64–0.92; P=0.005), the GP seeing more than 11 paediatric patients per week (RR 0.72; 95% CI 0.56–0.92; P=0.009) and the GP reporting that they were influenced in their decision to refer by the belief that a paediatrician would better manage the child’s condition ‘sometimes’ (RR 0.56; 95% CI 0.39–0.80; P= 0.002) or ‘frequently’ (RR 0.60; 95% CI 0.42–0.87; P=0.006).

Figure 1. Mean risk ratios (RRs) and 95% confidence intervals (CIs) for clinic-, general practitioner (GP)- and child-level factors associated with onward referral at clinic visits. Click here to enlarge.

IRSAD, Index of Relative Socio-economic Advantage and Disadvantage; Paed, paediatrician; SEIFA, Socio-Economic Indexes for Areas.

Aim 4: Cost associated with each referral type, split by reason for initial visit

Estimated costs associated with each reason for visit category and by type of referral made are presented in Table 4. The total cost of referrals was estimated as $1.39 million, equating to a mean cost of $297.96 for each patient referred and a mean cost of $32.03 per patient when considering all patients seen during the study period. The largest estimated referral costs came from children referred to a private specialist ($837,221; 60.4% of total cost). When the take-up of referrals was assumed to decrease from 100% to 80%, the total estimated cost of referrals decreased to $1.08 million, and at 60% uptake it decreased further to $784,142.

Discussion

This is the first study to describe the paediatric case mix seen by Australian GPs, their referral patterns, factors associated with referrals and costs. Across 49,932 consultations, GPs saw children most commonly for medical conditions, immunisations and check-ups. Overall, GPs referred 10% of patients. The factor most strongly associated with referral was practice private billing. The total cost of referrals was estimated as $1.39 million (ie an average of $297.96 per patient referred). The largest category of referral cost was ‘private specialists’, predominantly paediatricians.

Mental health or developmental–behavioural problems were commonly referred and usually to private specialists. Australian GP registrars have reported a relative lack of confidence in managing these sorts of problems compared with medical problems in children, but it is unknown whether the same is true for practising GPs.¹⁰ The fact that private-billing practices were more likely than bulk- or mixed-billing practices to refer children is potentially concerning. It could be due to selection bias, with parents attending private-billing practices more likely to request referrals to a private paediatrician, or due to inequity of referrals. Children from lower socioeconomic areas are more likely to have mental health and developmental–behavioural problems than those from wealthier areas.¹¹ However, these children, being more likely to attend bulk billing practices, received fewer referrals for specialist care. Although female GPs spend longer in consultations than male GPs,¹² it is unclear why being a female GP was associated with more referrals, even after controlling for GP case mix by gender. This requires further investigation.

GPs who had been practising for 6–15 years were the least likely to refer. This could reflect a ‘sweet spot’ whereby GPs with more experience than those working for less than six years have greater confidence to manage mental health and developmental–behavioural problems, whereas those practising for longer might have missed out on training in the ‘new morbidities’ of developmental–behavioural and mental health problems.¹³

This study has several strengths. We used objective measures of outcomes from general practice EMRs and included rich data from practices with a range of practice billing types, GP experience and patient socioeconomic status. The study also has some limitations. We sampled metropolitan practices only, so future research should examine rural practices. Practices self-selected, so might represent a bias towards paediatric care, but participating practices were similar in socioeconomic status to other practices in their primary health network. We used EMR data entered by GPs, and thus some data might be incomplete or incorrectly entered. GPs sometimes recorded several reasons for a visit, but for the purposes of this analysis, we only coded the first reason entered. However, we believe that basing our analyses on the first reason for the visit recorded in the EMR allows for an understanding of the broad patterns of referrals and associated factors and costs, given that anecdotally (we could find no published evidence) clinicians tend to enter the main diagnosis first. Finally, we collected data during the COVID-19 pandemic and associated lockdowns when paediatric ED presentations for infections and injuries declined and fewer children visited GPs.^14,15 This will have affected referral patterns.

Conclusion

In conclusion, Australian GPs see children for a range of conditions. They are more likely to refer mental health and developmental–behavioural than other conditions. Models of care that increase the equity of referrals for such conditions (eg integrated primary and specialist care,⁵ school-based healthcare^16,17) are crucial if we are to provide specialist care to all Australian children who need it. Paediatric referrals incur a significant cost, which presents an important opportunity for strengthening primary care to be more equitable and efficient.