The quality of preventive care for pre-school aged children in Australian general practice

Background Variable and poor care quality are important causes of preventable patient harm. Many patients receive less than recommended care, but the extent of the problem remains largely unknown. The CareTrack Kids (CTK) research programme sought to address this evidence gap by developing a set of indicators to measure the quality of care for common paediatric conditions. In this study, we focus on one clinical area, ‘preventive care’ for pre-school aged children. Our objectives were two-fold: (i) develop and validate preventive care quality indicators and (ii) apply them in general medical practice to measure adherence. Methods Clinical experts (n = 6) developed indicator questions (IQs) from clinical practice guideline (CPG) recommendations using a multi-stage modified Delphi process, which were pilot tested in general practice. The medical records of Australian children (n = 976) from general practices (n = 80) in Queensland, New South Wales and South Australia identified as having a consultation for one of 17 CTK conditions of interest were retrospectively reviewed by trained paediatric nurses. Statistical analyses were performed to estimate percentage compliance and its 95% confidence intervals. Results IQs (n = 43) and eight care ‘bundles’ were developed and validated. Care was delivered in line with the IQs in 43.3% of eligible healthcare encounters (95% CI 30.5–56.7). The bundles of care with the highest compliance were ‘immunisation’ (80.1%, 95% CI 65.7–90.4), ‘anthropometric measurements’ (52.7%, 95% CI 35.6–69.4) and ‘nutrition assessments’ (38.5%, 95% CI 24.3–54.3), and lowest for ‘visual assessment’ (17.9%, 95% CI 8.2–31.9), ‘musculoskeletal examinations’ (24.4%, 95% CI 13.1–39.1) and ‘cardiovascular examinations’ (30.9%, 95% CI 12.3–55.5). Conclusions This study is the first known attempt to develop specific preventive care quality indicators and measure their delivery to Australian children in general practice. Our findings that preventive care is not reliably delivered to all Australian children and that there is substantial variation in adherence with the IQs provide a starting point for clinicians, researchers and policy makers when considering how the gap between recommended and actual care may be narrowed. The findings may also help inform the development of specific improvement interventions, incentives and national standards.


Background
Preventive care is important for optimising every child's growth and development, especially in the early years [1][2][3][4][5]. While health outcomes for Australian children concord with international studies from the USA [6,7], the underlying processes and reliability of care are unknown. That is, we do not know whether, or to what extent, evidence-and consensus-based recommended care is actually delivered. Past research suggests that preventive care services afford many benefits to children, including improved healthcare utilisation patterns [8,9], better parental assessment of child health [8,9], increased adherence to health-promoting behaviours [9] and enhanced family functioning [8].
In Australia, approximately 5% of all general practitioner (GP) consultations per year are for children aged under 5 years [10]. While a number of different primary care clinicians are available to children and their parents (e.g. child and family health nurses) [11], GPs provide the majority of primary care services for families and children from birth to school entry [12]. This means GPs are ideally placed to play a central role in the early detection of developmental and behavioural problems in children, for disease prevention and health promotion [13].
There is evidence that the care delivered to children is not optimal and varies widely between conditions and healthcare providers. The study by Mangione et al. [7] reported that in the years 1996 to 2000, children received recommended care 68% of the time for acute medical problems, 53% for chronic medical conditions and 41% for preventive healthcare, with an overall average of 47%. Long-term negative sequelae from inadequate preventive screening and care services have been well documented [14][15][16].
Comprehensive national-level interventions aimed at implementing clinical practice guidelines (CPGs) for preventive child healthcare have produced mixed results; while improvements in general uptake were seen, sustained effects were hampered by issues with adherence to individual core elements, such as making and tracking referrals [17], systematically screening for psychosocial problems and visual disorders [18] and a lack of reliable and feasible metrics [18]. Before future implementation and evaluation of the effect of updated CPGs on clinical practice in Australia, information at a population level regarding whether, and to what extent, recommended healthcare is delivered for children for a range of conditions is needed [6,19].
The CareTrack Kids (CTK) programme of research sought to address this evidence gap by developing a set of care quality indicators for a range of common paediatric conditions and measuring the frequency with which they were delivered. CTK aimed to assess the quality of care Australian children aged 0-15 years received in 2012 and 2013 by measuring adherence to evidence and consensus-derived recommendations for 17 important common conditions (Table 1) [6].
A separate aim of CTK was to determine the quality of preventive care pre-school aged children received, which is the focus of this study. Our objectives were two-fold: (i) develop and validate preventive care quality indicators and (ii) apply them across GP settings, to measure adherence.

Methods
The broader methods of the CTK research programme have been described before [6,19,20]. This section provides further information specific to this study about preventive care. A more detailed description is provided as a supplementary file (Additional file 1).

Study design, sample and setting
A retrospective review of a population-based sample of children's general practice medical records was undertaken. We assessed preventive care delivered in 2012 and 2013 to children aged from 2 months to (and including) 4 years by GPs located in selected health department Health Districts (administrative units through which local health departments deliver health services) across three Australian states (Queensland, New South Wales and South Australia) [6]. Clinical indicators, used to assess the quality of preventive care delivered, were developed by expert panels using a modified Research ANd Development- University of California Los Angeles (RAND-UCLA) consensus-building Delphi process [20,21].
A sample size for this study was not pre-specified, as preventive care was sampled within medical records that contained occasions of care for one or more of the 17 CTK conditions. Within the selected Health Districts, we advertised the study to GP clinics (making contact with GPs and practice managers) and approached all the providers we could identify through internet searches and via personal contacts. We sampled medical records from the participating general practices for occasions of care for different subsets of eight to nine of the 17 CTK conditions. Further details of our sampling strategy are described in Additional file 1. We estimate that this sampling method covered conditions responsible for approximately 40% of GP consultations with children. Five South Australian GP clinics (located in the same Australian state as the CTK project team members) were recruited for the pilot study. These clinics did not participate in the main study and data were used only for testing and refining methodological processes and the practical application of the indicators; pilot data were not included in the analyses or findings of the main study.
Preventive care was sampled from medical records which were selected because they contained a consultation for one of 17 targeted CTK medical conditions. There are two implications of this approach; firstly, consultations for each condition had different age structures, resulting in different numbers of children born in different periods, ranging from approximately 100 born in 2013 to around 300 in other years. Secondly, the different follow-up periods mean that we sampled proportionately more children at some ages, because these were designated longer follow-up windows (see Additional file 1, section 1.3). Both issues, which manifested as different numbers of children from different age groups being included in the sample (potentially overor under-representing some), were addressed in analysis by using appropriately calculated weights (see Additional file 1, section 1.5).

Development and validation of clinical indicators and indicator questions (IQs)
For the purposes of this study, a clinical indicator was defined as 'a measurable component of a standard or guideline, with explicit criteria for inclusion, exclusion, time frame and practice setting' [20]. We developed and validated a set of care indicators in a consecutive sixstep process [22]: 1. The national and international literature were searched for CPGs about paediatric preventive care [20], and one CPG was found [23].
2. Seventy-one recommendations about specific aspects of care delivery were extracted from the CPG with their associated grades of evidence. 3. The recommendations were reduced to 24 by excluding those that were descriptive statements rather than specific actions; used the terms 'may', 'consider' and 'could'; unlikely to be documented in a medical record; or described organisational or systems-based rather than patient-level actions. 4. An expert group of clinicians from the CTK team, consisting of two GPs and one paediatrician, assessed and refined the remaining recommendations during three rounds of the modified RAND-UCLA Delphi process. 5. The draft indicators were assessed and validated by an expert panel of clinicians (two GPs and a paediatrician) who were not part of the CTK team (including definitions for the criteria against which compliance could be measured). 6. The final clinical indicators (n = 7) were phrased as specific indicator questions (IQs) (n = 43) that could be applied during record review and pilottested in general practice.
The expert clinicians (n = 6) were recruited through advertisements in medical colleges and professional associations and networks. They assessed the recommendations using a modified RAND-UCLA method for scoring, which is a recognised approach to derive indicators with content, construct and predictive validity in a reliable manner [21]. The experts indicated whether the recommendations and indicators were acceptable (i.e. reflected 'essential' Australian clinical practice in 2012-2013); feasible to detect in medical records; and of potential clinical value [20]. In addition to these three criteria, the experts who were not part of the CTK team (step 5), scored indicators on a 9-point Likert scale according to their representativeness of recommended care delivered to children during 2012 and 2013. All experts also recommended for exclusion indicators which duplicated concepts covered in other indicators and provided written feedback (e.g. on the wording of indicators) throughout the process.

Data collection
Data were collected to assess eligibility for indicator assessment and compliance by eight paediatric nurses (surveyors) who were trained and passed competency evaluations via a 5-day programme. A surveyor manual was developed which included instructions, conditionspecific definitions, inclusion and exclusion criteria, and guidance for assessing eligibility of each encounter for relevant indicators. A web-based tool, originally developed for the CareTrack Australia adults study [24,25] was modified to enter data during the medical record review, which included embedding algorithms to filter indicators by age. As participating general practices were separated by as much as 3000 km (2000 miles), assessing interrater reliability using the actual medical records was not feasible. For the overall CTK analysis, mock medical records were therefore assessed and compared for 6 of the 9 surveyors, and a good level of agreement was found for children's eligibility for indicator assessment (K = 0.76; 95% CI 0.75-0.77; n = 1895) and assessment of indicators (K = 0.71; 95% CI 0.69-0.73; n = 1009). Specifically, for the preventive care mock records, agreement was deemed fair for indicator eligibility (k = 0.49; 95% CI 0.41-0.57; n = 42) and assessment (k = 0.69; 95% CI 0.51-0.86; n = 42).

Analysis
At indicator level, estimates of compliance were measured as the percentage of eligible indicators (i.e. indicators answered either 'Yes' or 'No') which were scored as 'Yes'. Compliance results for some clinically related indicators were aggregated as bundles of care. Two different types of bundles were created: The weighted data were analysed in SAS v9.4, using the SURVEYFREQ procedure. Variance was estimated by Taylor series linearization. State was specified as a stratum, and the primary sampling unit (Health District) was specified as the clustering unit. Exact 95% CIs were generated using the modified Clopper-Pearson method. The overall estimate of recommended preventive care was the weighted average of the individual indicator assessments.

Ethical considerations
The study was approved by the Royal Australian College of General Practitioners (NREEC 14-008). Australian Human Research Ethics Committees can waive requirements for patient consent for external access to medical records if the study entails minimal risk to healthcare providers and patients [19]; all relevant bodies provided this approval. Participants were protected from litigation by gaining statutory immunity for CTK as a quality assurance activity, from the Federal Minister for Health under Part VC of the Health Insurance Act 1973 (Commonwealth of Australia).

Indicators
Our development and validation process resulted in seven clinical indicators, which were formatted into 43 medical record audit IQs. The IQs are listed in Table 3, with additional details in Additional file 2. The 43 IQs were grouped into six age group care bundles (Table 4) and eight clinical care bundles (Table 5).

Indicator assessments
Of 132,268 possible indicator assessments, 125,561 (94.9%) were automatically filtered out because of age restrictions and an additional 871 (0.7%) were designated as not applicable or ineligible. The surveyors undertook 976 record reviews and 5836 eligible indicator assessments in 80 GP practices. Additional file 3 provides definitions for the criteria against which surveyors marked indicator compliance.

Characteristics of reviewed records
The individual records of 976 children with one or more preventive care indicators were reviewed ( Table 2). Of these, 515 (52.8%) were male. The mean number of indicators was six per record (range 1 to 30).

Quality of preventive care
The assessed quality of treatment for each IQ is shown in Table 3

Key findings
This is the first known attempt to develop and validate preventive care quality indicators and apply them in general medical practice to measure adherence. We found on average that less than half of all indicator assessments resulted in the provision of recommended preventive care. When assessed as clinical care bundles, estimated GP performance was above this average when providing immunisations and undertaking anthropometric measurements, and below for assessments of nutrition, parental concerns, developmental progress and behaviour, cardiovascular status, the musculoskeletal system and vision. Estimated adherence to CPGs was highest for the complete bundle of preventive care delivered to 24month-old children and lowest for those aged 4 months.

Comparative performance
Our findings are broadly consistent with similar studies which assessed recommended preventive care for adults in Australia (42%) [24] and the USA (54.9%) [26]. Despite considerable efforts to promote and facilitate the uptake of CPGs in Australia [2,6], and the importance of early detection of disease with associated reductions in morbidity and mortality, paediatric preventive care is in line with CPGs less than half the time in both developed and developing health systems [27].
In our study, for the years 2012-2013, recommended preventive care was delivered to Australian children less often than the overall estimate of 59.8% across 17 common medical conditions in the CTK study [6]. It was similar to that reported for preventive care services delivered to American children between 1996 and 2000 (38.3% 'well-child care') [7].
Documented compliance with the 'visual assessment bundle' was found in less than one in five records. This is the first known measure of paediatric preventive eye care and a potentially concerning finding, given that one in ten Australian children aged 0 to 14 years has a vision disorder [28]. Recommended musculoskeletal examinations also demonstrated low compliance in our CTK population (24.4%; 95% CI 13.1-39.1). While the prevalence of musculoskeletal conditions in children aged less than 4 years of age who present to Australian primary care is estimated at less than 2% [29], given the volume of paediatric consultations in general practice, this nevertheless represents a significant number of children [30]. A complete musculoskeletal assessment has additional benefits as it may predict or detect other conditions in young children early to facilitate prompt initiation of treatment, such as cerebral palsy [31] and torticollis [32].
Documented compliance with recommended developmental progress and behaviour examinations in our study was substantially less than that reported in related literature. For example, assessment of developmental milestones ranged from 49.9% (no electronic medical record or structured well-baby visit record) to 80.3% (use of the Rourke Baby Record (RBR)) [33] for children aged up to 22 months, and 68% in a study where paediatric primary care practices collected their own data [17]. While not directly comparable, the findings suggest lower CTK compliances may be due to differing age  distributions of sample populations [33] and clinical contexts, where the imminent commencement of a 9month intervention to implement American Academy of Pediatrics recommendations for developmental screening and referrals may have raised awareness of the problem [17]. Discrepancies in compliance with recommended nutrition assessments (38.5% in CTK, 69.8% in [33]) may reflect disparate sample age groups.
Documented compliance was highest with the 'immunisation' bundle, with approximately four in five children receiving the recommended care. This is higher than rates reported in two studies from the USA that used comparable data collection methods. In a 2007 study [7], only 49.8% of children were fully immunised, but CTK did not follow infants longitudinally. In a 2015 study [34], it was reported that 78.4% of newborns and 56% of infants and toddlers were up-to-date with their immunisation status. On the other hand, higher rates of immunisation coverage have been reported in other primary care settings, including New Zealand (95.8%) [35] and Belgium (96.6-99.5%) [36]. The variation in immunisation rates between studies might be because of sampling considerations (i.e. a single-site vs the CTK population-based survey) [35], different data sources (e.g. electronic vaccine ordering and registration system vs CTK medical record audits) [36], specific vaccinations [36] and true inter-country differences [34,36]. Another consideration may be that our CTK study did not formally evaluate the timeliness of vaccination [35,36]; while a range of age groups (ages 2, 4, and 6 months and 2 and 4 years) were covered by our set of IQs, we did not specifically gather data on compliance with immunisation schedules for children between the ages of 6 months and 2 years, or those aged 3 years. This means that if CTK children had experienced delays in their immunisation but vaccines were complete by the ages of 2 or 4 years, our IQs still considered vaccinations to have been delivered on schedule [37].

Implications and next steps
Barriers to the provision of well-child healthcare have been proffered by Australian GPs and include time constraints, the financial status of families in lower socioeconomic groups who may not be able to cover gap payments, a lack of knowledge around the availability of and access to paediatric services, and fragmented care as a result of poor interflow of information among healthcare providers [2]. Shortfalls in the financial recompense available to GPs who provide preventive paediatric or 'well-child' care may be an additional barrier [38]; the Healthy Kids Check [39] which was operational during the study period enabled a national government rebate (Medicare) to be claimed by GPs only when a child was 4 years of age and had an up-to-date patient history, a complete immunisation schedule and been given a health promotion booklet [40]. However, the Check was retired in 2015 due to underperformance, cost blowouts and duplication of state and territory-based programs [41]. There are many potential interventions and initiatives to address barriers and improve preventive care [42], but success is variable, context-dependent and, too often, not rigorously assessed [43][44][45][46][47][48][49]. It is unlikely that any single intervention aimed at improving quality of care will yield significant or sustained benefits, and there is a need to engage, empower and support parents of   (13.4, 26.9) children with knowledge of the importance, and availability of, resources for preventive health in early childhood to help to optimise preventive care for young children [2,42,50,51]. Two of the key requirements for future interventions are therefore standards and reliable metrics-both components are potentially provided by this study. Lack of agreed definitions and established clinical standards of what constitutes recommended (or quality) preventive care is a barrier to best practice. Internationally agreed standards on well-child management and preventive care would allow comparison between healthcare systems and facilitate evaluation of the effectiveness of interventions. The Royal Australian College of General Practitioners (RACGP) have produced a national CPG for preventive activities in children and young people [52]. The development and implementation of guideline recommendations as measurable standards is strongly advocated as a means by which to initially audit and then ultimately improve quality of care [2,24,53].
A way forward could be the use of integrated e-health medical records embedded with clinical standards to help facilitate the provision of recommended care, consistency and completeness of documentation and enable large-scale surveillance or regionally based audits of current practice as well as for responsiveness to national initiatives to improve healthcare [24]. Future initiatives and interventions must target practitioners and consumers (paediatric patients and their parents) to optimise individual patient outcomes in addition to process measures.

Strengths and limitations
The key strength of the CTK study is that it was designed to be representative of a broad segment of the Australian population rather than a convenience or purposive sample. However, we did not collect sociodemographic data or randomly select all medical records to be reviewed; records were only sampled if they were identified as having a consultation for one of 17 common paediatric conditions covering approximately 40% of GP consultations with children. This sampling method could bias our estimates, if the sampled conditions are correlated with the quality of preventive care. Therefore, our results are only strictly generalisable to children who have one or more of the 17 conditions sampled in CTK at some point in the 2-year period of interest. This approach was used, rather than selecting another random sample of children without one of the 17 CTK conditions, to minimise the workload on administrative staff at participating general practices who were tasked with generating de-identified lists of eligible children and facilitating surveyor access of their records.
We included only GP records and occasions of service (including care provided by general practice nurses). Our findings therefore do not reflect the quality of preventive care delivered by other types of primary care clinicians. For example, children receiving immunisations from community child-health clinic nurses were not captured in this study as their treatment may not have been documented in general practice records [54]. In Australia, GPs are the predominant provider of paediatric vaccination services delivering nearly three-quarters of all vaccinations for children aged 0-6 years, which is in addition to being regularly contracted by State Governments to deliver the childhood vaccination programs provided through schools and community centres [55]. An investigation of the quality of preventive care provided to older children would be an important focus of future research as no other primary care clinicians provide well-child care for this age group. Integrated ehealth medical records which record both GP and 'maternal and child health' data would facilitate the feasibility of such an endeavour. The IQs were drawn from one CPG relevant to Australian general practice in 2012-2013, which limits the applicability of the IQs to other contexts and settings. Six clinical experts (three within the CTK team and three external to it) participated in the Delphi process to develop the IQs. This may have adversely affected the face validity of 'preventive care' IQs. The IQs, which were developed and validated using rigorous methods to measure compliance, measure most, but not all, aspects of preventive care for children aged from 2 months to 4 years. Age group bundles will have lower compliance scores than clinical care bundles because the former are linked by 'AND' statements; compliance with an age group bundle cannot be higher than the compliance for the IQ with the lowest compliance. Clinical care bundles, on the other hand, are weighted averages of all the component indicators assessed. The use of bundles as composite measures in addition to individual IQs provide an alternative perspective on care quality [56].
Our findings provide a snapshot in time, and compliance rates may have changed since 2013. Despite considerable investment in developing and disseminating comprehensive clinical practice guidelines [52], the latest Australian Research Alliance for Children and Youth (ARACY) report card suggests that some outcomes (e.g. immunisation against measles and whooping cough) have worsened since the last report in 2013 [57]. The CTK study has relied on process (rather than outcomebased) indicators to audit care as documented within the medical record. Previous research has suggested that medical records may underestimate the quality of care provided by healthcare services, with doctors being more likely to document some aspects of care (e.g. medication, immunisations) than others in medical records (e.g. patient history information provided to patients) [58,59]. The record review method depends on the quality of clinical records and surveyor characteristics; however, we sought to pre-emptively address this by using a twostage structured record review method, training and testing surveyors prior to data collection and providing a coding manual.

Conclusion
Our findings that preventive care is not reliably delivered to all Australian children and that there is substantial variation in adherence with the indicators provide a starting point for clinicians, researchers and policy makers when they consider how the gap between recommended and actual care may feasibly be narrowed. The findings may also help inform the development of specific improvement interventions, incentives and national standards.