The effect of different methods of remuneration on the behaviour of primary care dentists
Abstract
Background
Methods of remuneration have been linked with the professional behaviour of primary care physicians. In dentistry, this can be exacerbated as clinicians operate their practices as businesses and take the full financial risk of the provision of services. The main methods for remunerating primary care dentists include fee‐for‐service, fixed salary and capitation payments. The aim of this review was to determine the impact that these remuneration mechanisms have upon primary care dentists’ behaviour.
Objectives
To evaluate the effects of different methods of remuneration on the level and mix of activities provided by primary care dentists and the impact this has on patient outcomes.
Search methods
We searched the Cochrane Effective Practice and Organisation of Care (EPOC) Group Specialised Register; the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 7, 2013); MEDLINE (Ovid) (1947 to 11 June 2013); EMBASE (Ovid) (1947 to 11 June 2013); EconLit (1969 to 11 June 2013); the NHS Economic Evaluation Database (EED) (11 June 2013); and the Health Economic Evaluations Database (HEED) (11 June 2013). We conducted cited reference searches for the included studies in ISI Web of Knowledge; searched grey literature sources; handsearched selected journals; and contacted authors of relevant studies.
Selection criteria
Primary care dentists were defined as clinicians that deliver routine or mainstream dental care in a primary care environment. We included randomised controlled trials (RCTs), non‐randomised controlled clinical trials (NRCTs), controlled before‐after (CBA) studies and interrupted time series (ITS) studies. The methods of remuneration that we considered were: fee‐for‐service, fixed salary and capitation payments. Primary outcome measures were: measures of clinical activity; volume of clinical activity undertaken; time taken and clinical session length, or both; clinician type utilised; measures of health service utilisation; access and attendance as a proportion of the population; re‐attendance rates; recall frequency; levels of oral health inequalities; non‐attendance rates; healthcare costs; measures of patient outcomes; disease reduction; health maintenance; and patient satisfaction. We also considered measures of practice profitability/income and any reported unintended effects of the included methods of remuneration.
Data collection and analysis
Three of the review authors (PRB, JP, AMG) independently reviewed titles and abstracts and resolved disagreements by discussion. The same three review authors undertook data extraction and assessed the quality of the evidence from all the studies that met the selection criteria, according to Cochrane Collaboration procedures.
Main results
Two cluster‐RCTs, with data from 503 dental practices, representing 821 dentists and 4771 patients, met the selection criteria. We judged the risk of bias to be high for both studies and the overall quality of the evidence was low/very low for all outcomes, as assessed using the GRADE approach.
One study used a factorial design to investigate the impact of fee‐for‐service and an educational intervention on the placement of fissure sealants in permanent molar teeth. The authors reported a statistically significant increase in clinical activity in the arm that was incentivised with a fee‐for‐service payment. However, the study was conducted in the four most deprived areas of Scotland, so the applicability of the findings to other settings may be limited. The study did not report data on measures of health service utilisation or measures of patient outcomes.
The second study used a parallel group design undertaken over a three‐year period to compare the impact of capitation payments with fee‐for‐service payments on primary care dentists’ clinical activity. The study reported on measures of clinical activity (mean percentage of children receiving active preventive advice, health service utilisation (mean number of visits), patient outcomes (mean number of filled teeth, mean percentage of children having one or more teeth extracted and the mean number of decayed teeth) and healthcare costs (mean expenditure). Teeth were restored at a later stage in the disease process in the capitation system and the clinicians tended to see their patients less frequently and tended to carry out fewer fillings and extractions, but also tended to give more preventive advice.
There was insufficient information regarding the cost‐effectiveness of the different remuneration methods.
Authors' conclusions
Financial incentives within remuneration systems may produce changes to clinical activity undertaken by primary care dentists. However, the number of included studies is limited and the quality of the evidence from the two included studies was low/very low for all outcomes. Further experimental research in this area is highly recommended given the potential impact of financial incentives on clinical activity, and particular attention should be paid to the impact this has on patient outcomes.
PICOs
Plain language summary
The effect of different methods of remuneration on the behaviour of primary care dentists
Financial incentives within remuneration systems (methods of payment) can influence the behaviour of clinicians working in primary care environments. Systematic reviews in medicine have found that changing the way that doctors are paid can produce substantial changes in the types of activities that are undertaken. For example, paying a fee for specific services can increase the quantity of services delivered, although this may not produce an improvement in patient outcomes.
The main methods for remunerating primary care dentists include:
1. fee‐for‐service payment (a payment made to a dentist for every item of service or unit of care that they provide);
2. fixed salary payment (a lump sum payment made to a dentist for a set number of working hours or sessions per week);
3. capitation payment (a payment based on the number and types of patients whose care the dentist takes responsibility for); and
4. blended payment (combination of above).
Our review identified two studies examining the effects of different methods of remuneration on the behaviour of 821 dentists from 503 dental practices, involving 4771 patients. Both were conducted in the United Kingdom. One study investigated the impact of a fee‐for‐service payment and an educational intervention on the placement of fissure sealants in permanent molar teeth. The second study compared the impact of capitation payments and fee‐for‐service payments on primary care dentists’ clinical activity and the levels of dental decay that were experienced across the two payment systems.
The first study found an increase in clinical activity related to fee‐for‐service payments. In the second study, dentists working under capitation arrangements restored carious teeth at a later stage in the disease process than fee‐for‐service controls. In the capitation arm, the dentists tended to see their patients less frequently and tended to carry out fewer fillings and extractions, but tended to give more preventive advice.
There was insufficient information regarding cost‐effectiveness of the different remuneration methods.
Financial incentives within remuneration systems may produce changes to clinical activity undertaken by primary care dentists. However, the number of included studies is limited and the quality of the evidence is low/very low for all outcomes.
Authors' conclusions
Summary of findings
| Population: Dentists seeing children with erupted second permanent molars Control: No specific intervention | |||||
| Outcomes | Illustrative comparative risks (95% CI) | Relative effect | No. of Participants | Quality of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Control | Fee‐for‐service remuneration | ||||
| Measures of clinical activity: mean percentage of 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars per dentist (weighted by number of children seen) ‐ adjusted1 | ‐ | RD 9.8% higher (1.8% higher to 17.8% higher)3 | ‐ | 133 dentists | ⊕⊝⊝⊝ |
| Measures of clinical activity: mean percentage of 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars per dentist (weighted by number of children seen) ‐ unadjusted2 | 26.3% (CI NR) | RD 7.1% higher (1.9% lower to 16.1% higher) | RR 0.27 (CI NR) | 133 dentists | ⊕⊝⊝⊝ very low4 |
| Healthcare costs: cost‐effectiveness of fee‐for‐service vs. control (reported as the "% change in outcome per £[GBP]" ‐ currency year NR)5 | ‐ | ‐ | 0.10 (CI NR) | 68 dentists | ⊕⊝⊝⊝ very low4 |
| CI: Confidence interval; GBP: Pound Sterling; NR: Not reported; RD: Risk difference; RR: Risk ratio | |||||
| GRADE Working Group grades of evidence | |||||
| 1The model adjusted for the baseline dental practice‐level covariates (deprivation category for the area of dental practice, number of partners in practice, throughput of 11‐ to 13‐year‐olds and the number of restorative fissure sealants placed on first permanent molars at baseline). 2The basis for the assumed risk is the risk in the control group (i.e. the probability of a dentist in the control group fissure sealing a second permanent molar of a 12‐ to 14‐year‐old). The corresponding risk (the risk difference) is based on the assumed risk in the control group and the relative effect of the fee‐for‐service remuneration (the risk ratio). 3Statistically significant at the 5% level.
5Outcome in the incremental cost‐effectiveness ratio (ICER) is not specified and can only be assumed to relate to sealant placement | |||||
| Population: Children undergoing routine continuing dental care at 354 dental practices (the number of dentists varied over time as dentists joined and left dental practices, so the number of dental practices was the stable and primary parameter; the total number of 0‐ to 15‐year‐old children was not reported accurately) | |||||
| Outcomes | Illustrative comparative risks (95% CI) | Relative effect | No. of participants | Quality of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Fee‐for‐service remuneration | Capitation remuneration | ||||
| Measures of health service utilisation: mean number of visits per 0‐ to 15‐year‐old | Northern urban community: 2.8 | MD 0.4 lower1 | ‐ | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 2.3 | MD 0.4 lower1 | ||||
| Rural community: 2.5 | MD 0.3 lower | ||||
| Scottish community: 2.8 | MD 0.5 lower1 | ||||
| Patient outcomes: mean number of filled teeth per 0‐ to 15‐year‐old | Northern urban community: 0.78 | MD 0.18 lower1 | ‐ | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 0.34 | MD 0.03 lower | ||||
| Rural community: 0.44 | MD 0.21 lower | ||||
| Scottish community: 0.91 | MD 0.28 lower1 | ||||
| Patient outcomes: mean percentage of 0‐ to 15‐year‐olds having one or more teeth extracted3 | Northern urban community: 18% | RD 5% lower1 | RR 28% lower1 | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 7% | RD 0% | RR 0% | |||
| Rural community: 10% | RD 3% lower1 | RR 30% lower1 | |||
| Scottish community: 15% | RD1% lower | RR 7% lower | |||
| Patient outcomes: mean number of decayed teeth per 14‐ to 15‐year‐old (data for 0‐ to 15‐year‐olds NR) | Northern urban community: 0.16 | MD 0.16 higher | ‐ | 1919 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 0.24 | MD 0.07 higher | ||||
| Rural community: 0.58 | MD 0.75 higher2 | ||||
| Scottish community: 0.65 | MD 0.15 higher | ||||
| Measures of clinical activity: mean percentage of 0‐ to 15‐year‐olds receiving active preventive advice3 | Northern urban community: 19% | RD 27% higher1 | RR 142% higher1 | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 18% | RD 15% higher | RR 83% higher | |||
| Rural community: 34% | RD 5% lower | RR 15% lower | |||
| Scottish community: 28% | RD 9% higher | RR 32% higher | |||
| Healthcare costs: mean expenditure in GBP (currency year NR) per 0‐ to 15‐year‐old5 | Northern urban community: 20.55 | MD 4.22 higher | 21% higher | 276,4145 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 16.67 | MD 6.18 higher | 37% higher | |||
| Rural community: 17.29 | MD 6.90 higher | 40% higher | |||
| Scottish community: 17.68 | MD 1.52 higher | 9% higher | |||
| CI: Confidence interval; GBP: Pound Sterling; MD: Mean difference; NR: Not reported; RD: Risk difference; RR: Risk ratio | |||||
| GRADE Working Group grades of evidence | |||||
| 1Statistically significant at the 5% level. However, the unit of analysis (e.g. dentists, patients, parents and administrators) was often not the same as the unit of randomisation. This leads to unit‐of‐analysis error, where P values are artificially small (though the estimates of effect are unbiased), leading to false positive conclusions that the intervention had an effect. 2Statistically significant at the 1% level. However, the unit of analysis (e.g. dentists, patients, parents and administrators) was often not the same as the unit of randomisation. This leads to unit‐of‐analysis error, where P values are artificially small (though the estimates of effect are unbiased), leading to false positive conclusions that the intervention had an effect. 3The basis for the assumed risk is the risk in the control group (i.e. the probability of a dentist in the control group giving preventive advice to or extracting a tooth for a 0‐ to 15‐year‐old). The corresponding risk (the risk difference) is based on the assumed risk in the control group and the relative effect of the capitation remuneration (the risk ratio). 4There were four matched pair of Health Service administrative areas. These randomised pairs were treated as separate, thus the overall study contained four replicates under contrasting socioeconomic and environmental circumstances. There were only two Health Service administrative areas randomised in each replicate, therefore each arm of each replicate only contained one Health Service administrative area. 5 All payments made to study dentists for the treatment of 0‐ to 5‐year‐olds during 1988 were divided by the estimated numbers of children treated. However, the estimated number is only an approximation as it was impossible to eliminate double‐counting, particularly in the fee‐for‐service system. This means that the mean expenditure per 0‐ to 15‐year‐old should only be considered as close approximations, and there is bias that places the capitation arm at a disadvantage. In addition, participating dentists in capitation areas referred significantly more children to the Community Dental Service compared to dentists in fee‐for‐service areas; this is despite the fact that non‐participating dentists in the capitation areas tended to refer significantly fewer children compared to non‐participating dentists in fee‐for‐service areas. The cost of treating the children in the Community Dental Service would not have been taken into account in the economic analysis. 6 Quality of the evidence
| |||||
Background
In medicine, methods of remuneration that form the provider payment have been linked with the clinical and professional behaviour of primary care physicians (Donaldson 1989).
In a Cochrane Effective Practice and Organisation of Care (EPOC) Group review of the effects of the method of remuneration on the behaviour of primary care physicians, fee‐for‐service payments were associated with an increase in the quantity of primary care services, but changes to patient outcomes were equivocal (Gosden 2000).
A more recent EPOC Group review examined the effect of financial incentives on the quality of health care provided by primary care physicians (Scott 2011). Again, there was insufficient evidence to determine the impact of financial incentives on the quality of primary health care. Six of the seven included studies demonstrated positive but modest effects on a minority of the measures of quality of care, and the remaining study found no effect. Most of the studies had a substantial risk of bias due to factors such as selection bias (due to non‐randomisation and, in randomised studies, due to analysis at the level of the medical group combined with lack of reporting on changes in the composition of these medical groups between baseline and follow‐up, or between the intervention and control groups).
An EPOC Group overview of reviews was carried out to evaluate the effects of financial incentives on the behaviour of healthcare professionals and patient outcomes (Flodgren 2011). This overview included four reviews, two of which were judged to be of moderate quality and the remaining two were judged to be of high quality. The 32 studies that these four reviews reported on were found to be of low to moderate quality. Fee‐for‐service and capitation payments were generally effective (improving 7/10 outcomes and 48/69 outcomes, respectively), while fixed salary payments were generally ineffective (improving 3/11 outcomes). The review also considered payments for providing a prespecified level of activity or providing a change in activity or quality of care, and found that this was generally effective (improving 17/20 outcomes). In addition, the review considered the effect of financial incentives in general across categories of outcomes and found that they were generally effective at improving processes of care, referrals and admissions, and prescribing costs. However, financial incentives were ineffective in improving compliance with guidelines' outcomes, and had mixed effectiveness on consultation rates. No evidence was found for the effect of financial incentives on patient outcomes. Vote counting was used to summarise the direction of the effect, rather than a meta‐analysis. Many studies utilised a controlled before‐after design, and adjusting for these reduced the overall impact on effectiveness. There were also concerns about the completeness and generalisability of the evidence.
Description of the condition
This review considered all aspects of dental care undertaken by primary care dentists (defined as clinicians that deliver routine or mainstream dental care in a primary care environment), excluding the provision of specialist services or the management of adult or child patients with special needs.
Description of the intervention
The main mechanisms for remunerating primary care dentists include fee‐for‐service, fixed salary and capitation payments (Grytten 2005) i.e. service throughput‐based (fee‐for‐service), time‐based (fixed salary) and patient‐based (capitation). These vary considerably across different countries and are heavily influenced by the prevailing political and professional culture (Grytten 2005). Capitation payments tend to secure effectiveness at the cost of patient selection and under treatment, while fee‐for‐service payments secure quality but often suffer from cost containment problems (Gosden 2000; Grytten 2005). For example, Birch 1988 found that where primary care dentists have a substantial influence over demand for care, there are strong incentives to over treat. Chalkley 2006 also found that treatment for patients exempt from payment was more intensive when provided by self employed primary care dentists compared to their salaried counterparts. In a natural experiment where public dental officers in one county were given the opportunity to renegotiate their contract from a fixed salary contract to a combined capitation and fixed salary contract, “the transition to an incentive‐based remuneration system led to an increase in the number of individuals under supervision, without either a fall in quality or a patient selection effect” (Grytten 2009). Salary remuneration removes the link between income and the level and type of services delivered, or patients served, leading to high costs per patient (Grytten 2005). A recent review of the impact of introducing a new National Health Service contract on the behaviour of primary care dentists in the United Kingdom found that clinicians were very sensitive to changes to remuneration (SDO 2011). This reduced job satisfaction and morale (Harris 2009), adversely affected patient access and changed the service and mix of activity, or led to a shift of primary care dentists from a national contract to the private sector (Steele 2009).
How the intervention might work
Financial incentives within health care remuneration systems have the potential to align the provision of health services with the aims of the health system, e.g. making the services more effective, more equitable or more patient‐centred. Financial incentives involve transferring money from ‘buyers’ (patients or third‐party payers such as governments or insurers) to ‘sellers’ (individuals or groups of clinicians, or their employers) on the condition that the sellers behave in a certain way, e.g. by providing a particular health service, sometimes at a specified level of quality (Scott 2011).
The economic theory explains that if the size of the payment is greater than the marginal (i.e. additional) costs of the behaviour change, then the cost‐benefit ratio of the behaviour change can be lowered and this can make the behaviour change more likely to happen (Scott 2011). There may be heterogeneity in the marginal costs of changing behaviour among the providers, e.g. due to differences in administrative costs of practices of different sizes, as larger practices may have lower unit costs (Scott 2011). The theory highlights that, in addition to the method of remuneration, other factors such as the level of payment (particularly in terms of the proportion of total revenue from the remuneration system) are likely to have an impact, due to the economic concepts of substitution and income effects. How payments are utilised by a practice is also significant, particularly regarding how the payments are distributed between groups of providers and whether any of the payments are invested into service provision to reduce the marginal costs (which reduces the cost‐benefit ratio), rather than being used to pay the providers (Scott 2011). The relative impact of other sources of motivation (such as professional autonomy) has a bearing on the effect of financial incentives, and these may vary for different providers and in different settings (Scott 2010; Scott 2011) . It is important to note that financial incentives may also influence the quality and cost of health service provision by influencing recruitment and retention and thereby influencing the mix of providers. If poorly designed, financial incentives can have unintended effects such as incentivising providers to prioritise one disease area at the expense of other disease areas, such that the overall net impact on health service provision is detrimental.
Primary care dentists operate their practices as businesses (Grytten 2005; Tickle 2011) so they differ from many other healthcare professionals in that they take all the financial risk for service provision, receiving little or no support to cover initial start‐up costs or for the development of their capital infrastructure. As a result, they are potentially more sensitive to financial incentives within the remuneration system, which represents their principal source of income; changes in the clinical activity of primary care dentists in the United Kingdom have been documented following the introduction of new methods of payment in the National Health Service (McDonald 2012; Tickle 2011). In addition, unlike primary care physicians, whose predominant function is the management of symptomatic patients or those with chronic conditions, the bulk of service delivery in dentistry in most industrialised countries, in terms of volume of activity, is based on the regular attendance of asymptomatic patients. This can produce distortions in both the demand and supply side of provision (Wright 2001). Demand can be influenced by health literacy and patient expectations of care (Gregory 2007; Milsom 2009; Steele 2009), while supply can be influenced by the financial incentives inherent within the remuneration system, leading to supplier‐induced demand (Birch 1988; Tickle 2011). Despite this, remuneration systems in primary care dentistry have received relatively little attention from a health economics perspective (Grytten 2009).
Why it is important to do this review
In dentistry, there is some evidence from observational study designs that methods of remuneration can impact on the behaviour of clinicians in primary care environments (Chalkley 2006; Grytten 2005; Tickle 2011). In medicine, fee‐for‐service payments are associated with an increase in the quantity of primary care services, but changes to patient outcomes are equivocal (Gosden 2000). As a result, it is important to understand the effects that different remuneration systems have on the pattern of service activity in dentistry and the patient outcomes generated (Grytten 2005). Evidence from experimental designs would also help to determine the most appropriate method of service delivery for the needs of a given population in order to inform future workforce planning (Grytten 2009).
Objectives
To evaluate the effects of different methods of remuneration on the level and mix of activities provided by primary care dentists and the impact this has on patient outcomes.
Methods
Criteria for considering studies for this review
Types of studies
We included the following study designs that met the Cochrane EPOC Group criteria (EPOC 2013).
-
Randomised controlled trials (RCTs)
-
Non‐randomised clinical trials (NRCTs)
-
Controlled before‐after (CBA) studies (at least two sites in each group)
-
Interrupted time series (ITS) studies
We reported numerical data on an individual study basis and outcome data for the multiple publications of one trial (Coventry 1989) were reported as one.
Types of participants
We examined studies involving primary care dentists providing routine dental care in primary care environments.
Types of interventions
We defined the method of remuneration as the payment that directly determines or influences the personal income of the primary care dentist. We included the following remuneration systems.
-
Fee‐for‐service payments
-
Fixed salary payments
-
Capitation payments
-
Blended payments (combinations of above)
Fee‐for‐service remuneration was defined as a payment made to a primary care dentist for every item of service or unit of care that they provide. Salaried remuneration was defined as a lump sum payment made to a primary care dentist for a set number of working hours or sessions per week. Capitation remuneration was defined as a payment based on the number and types of patients whose care the provider takes responsibility for.
Types of outcome measures
We only reported objective outcome measures and subjective outcome measures that used standardised validated instruments.
Primary outcomes
We considered the following as primary outcome measures.
-
Measures of clinical activity
-
Number of activities undertaken in a specified time period including examinations, oral hygiene instruction, scaling and polishing, periodontal treatment, restorations, root canal treatments, extractions and prostheses
-
Number of sessions over which treatment activity is distributed
-
Time taken and session length, or both, for treatment activities
-
Clinician type utilised
-
-
Measures of health service utilisation
-
Proportion of a population receiving care
-
Re‐attendance rates
-
Recall frequency
-
Levels of oral health inequalities by socio‐economic status, education or income
-
Proportion of population not receiving care (non‐attendance rates)
-
-
Healthcare costs
-
Patient outcomes
-
Disease reduction, including the number of new carious teeth, the proportion of patients with a basic periodontal examination greater than a score of two, and the proportion of patients with sites that bled on probing
-
Health maintenance, including the proportion of patients that did not require any operative treatment
-
Patient satisfaction, including the proportion satisfied with the dental care they received, the proportion satisfied with the waiting time for an appointment, and the proportion reporting that they felt involved in decisions about their care
-
Secondary outcomes
We considered the following as secondary outcome measures.
-
Measures of non‐clinical behaviour of primary care dentists including the rates of performing specified non‐clinical behaviours (e.g. education and training), when specified as a secondary outcome.
-
Measures of dental practice profitability/income.
-
Any unintended effects of the remuneration systems, including supplier‐induced demand when the service provided is not based on need (Birch 1988), changes to the types of treatment offered, and limitations to access (see Tickle 2011 for a conceptual framework).
Search methods for identification of studies
Michelle Fiander, Trials Search Co‐ordinator (TSC) for the Cochrane EPOC Group, wrote the search strategies. The TSC searched the Cochrane Database of Systematic Reviews and the Database of Abstracts of Reviews of Effects (DARE) for related systematic reviews, and the databases listed below for primary studies. Searches were conducted in June 2013; exact search dates for each database are included with the search strategies in Appendix 1.
Electronic searches
Databases
-
Cochrane Central Register of Controlled Trials (CENTRAL), Issue 7, 2013, Wiley
-
MEDLINE, 1946 ‐ June 2013, In‐Process and other non‐indexed citations, Ovid
-
EMBASE, 1947 ‐ June 2013, Ovid
-
EPOC Group, Specialised Register, June 2013, Reference Manager
-
EconLit, Dissertations & Theses, 1969 ‐ June 2013, ProQuest
-
PAIS International, Political Science,Worldwide Political Science Abstracts, June 2013, Proquest
-
CINAHL (Cumulative Index to Nursing and Allied Health Literature), 1980‐ June 2013, EbscoHost
-
NHS Economic Evaluation Database (EED), Issue 7, 2013, Wiley
-
Health Economic Evaluations Database (HEED), Issue 7, 2013, Wiley
We used search strategies that were comprised of keywords and, when available, controlled vocabulary such as MeSH (Medical Subject Headings). The TSC finalised search strategies using an iterative development process in which citations identified by various search terms were screened for relevance, either by review authors or the TSC. In this manner, individual terms and combinations of terms were assessed as relevant or irrelevant and were included or omitted from the final search strategies. We did not place any restrictions on either the date or language used. We searched all the databases from their start date forward.
We used two methodological search filters to limit retrieval to appropriate study designs: the Cochrane Highly Sensitive Search Strategy (sensitivity‐ and precision‐maximising version, 2008 revision) to identify randomised trials (Higgins 2011); and an EPOC methodology filter to identify non‐RCT designs. We have provided all the search strategies and specific run dates in Appendix 1.
Grey literature sources
We scanned publication titles on the following grey literature websites.
-
University of York (http://www.york.ac.uk/che/publications/)
-
University of Aberdeen, HERU (http://www.abdn.ac.uk/heru/publications/)
-
University of Sheffield (http://www.shef.ac.uk/scharr/sections/heds/discussion)
-
University of Bristol (http://www.bris.ac.uk/populationhealth/methodology/economics/)
-
Brunel University, HERG (http://www.brunel.ac.uk/about/acad/herg)
-
Swedish Institute of Health Economics (http://www.ihe.se/publiceringar‐1.aspx)
-
RAND Corporation (http://www.rand.org/pubs/research_briefs.html)
We examined websites for grey literature manually without using search interfaces as they do not usually support complex Boolean or other operators. We conducted the latest search in August 2013.
Searching other resources
We also undertook the following.
-
Reviewed reference lists of all included studies.
-
Conducted cited reference searches for all included studies in ISI Citation Indexes via Web of Knowledge.
-
Contacted authors of relevant studies/reviews to clarify reported published information and to seek unpublished results/data.
-
Screened the following individual journals from January 2010 to December 2012: Health Economics; Journal of Political Economy; Journal of Health Services Research and Policy; European Journal of Health Economics; and Journal of Applied Economics.
Data collection and analysis
We managed the whole review process using Review Manager 5 (RevMan 2012).
Selection of studies
After we had identified the titles and abstracts from the electronic searches, we downloaded them to a reference management database and removed the duplicates. Three of the review authors (PRB, JCP and AMG) independently examined the remaining references. We excluded studies that did not meet the inclusion criteria and obtained full‐text copies of the references that appeared to meet the inclusion criteria to assess for inclusion. We resolved differences by discussion and recorded the excluded studies in the Characteristics of excluded studies table.
Data extraction and management
Three of the review authors (PRB, JCP and AMG) independently extracted data from the included studies and resolved any differences by discussion.
We extracted the following data into the Characteristics of included studies tables.
-
Methods (study type and duration of study)
-
Participants (setting, unit of randomisation, unit of assessment/analysis, method of recruitment, inclusion criteria and exclusion criteria)
-
Interventions (details of interventions and control group)
-
Outcomes (primary and secondary (as specified in the protocol for this review) and adverse outcomes)
-
Source of funding
We extracted the following into the Appendices.
-
Number of participants (number randomised, number analysed and number not analysed with reasons, each by study arm)
-
Baseline characteristics and outcomes
Assessment of risk of bias in included studies
Three review authors (PRB, JCP and AMG) independently assessed the risk of bias of the included studies and considered other factors that affect the quality of evidence, including inconsistency, indirectness, imprecision and publication bias. We resolved disagreements by discussion.
We assessed the risk of bias for studies with a control group (RCTs, NRCTs and CBAs) using the following criteria (EPOC 2011; Higgins 2011).
-
Random sequence generation
-
Allocation concealment
-
Blinding of participants and personnel
-
Blinding of outcome assessment
-
Incomplete outcome data
-
Other bias (including baseline characteristics and outcomes, and protection against contamination)
We assessed ITS studies using the following criteria (EPOC 2011).
-
The intervention was independent of other changes
-
The shape of the intervention effect was prespecified
-
The intervention was unlikely to affect data collection
-
Incomplete outcome data were adequately addressed
-
The study was free from selective outcome reporting
We tabulated the description of the domains for each included study, along with a judgement on the risk of bias (low, high or unclear) for each domain, based on the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We planned to undertake a summary assessment of the risk of bias for the primary outcome across the studies (Higgins 2011). For each study, we provided the following summary assessment of the risk of bias.
-
Low risk when there is a low risk of bias across all domains.
-
Unclear risk of bias when there is an unclear risk of bias in one or more of the domains.
-
High risk of bias when there is a high risk of bias in one or more of the domains.
Measures of treatment effect
We reported outcomes for each included study in natural units.
For RCTs, NRCTs and CBAs, we reported pre‐intervention and post‐intervention means or proportions for all data points for both intervention and control groups where baseline results were available. Had data allowed, we had planned to calculate the unadjusted and adjusted (for any baseline imbalance) absolute change from baseline with 95% confidence intervals (CIs). For continuous variables, we reported mean differences (MDs). Dichotomous variables would have been reported as risk ratios (RRs) together with 95% CIs.
Had eligible ITS studies been identified, we would have extracted the difference in slope and the difference in pre‐ to post‐intervention levels. We had planned to analyse the post‐ versus pre‐intervention difference (adjusted for trends) at specific time points (three months, six months and six‐monthly thereafter). If the differences were not available in the primary reports, we would have reanalysed the results using data from graphs or tables.
We have presented the findings of the main comparisons from the included studies in the Summary of main results in order to interpret the results and draw conclusions about the effects of different interventions along with the quality of the evidence.
Unit of analysis issues
For cluster‐RCTs, we undertook analysis at the same level as the randomisation or at the individual level, accounting for the clustering. For cluster‐RCTs with unit of analysis error we did not report the P values or 95% CIs, as analyses not accounting for the design effect have the potential to inflate the type 1 error rate and result in artificially narrow CIs (Ukoumunne 1999). The point estimate is not affected by unit of analysis errors.
Dealing with missing data
We explicitly stated where studies had missing data.
Assessment of heterogeneity
We had planned to assess heterogeneity using The Cochrane Collaboration's test for heterogeneity, where P < 0.1 was to be considered significant (Higgins 2011). However, due to variations in comparisons made, plus methodological heterogeneity, it was felt inappropriate to pool data.
Assessment of reporting biases
If more than 10 studies had been identified for meta‐analysis, we had planned to assess publication bias according to the recommendations on testing for funnel plot asymmetry (Higgins 2011).
Data synthesis
We had planned to undertake meta‐analyses for clinically homogeneous RCTs that reported the same outcome measures: RRs for dichotomous data and MDs for continuous data, using random‐effects models (or fixed‐effect models if fewer than four studies were included). Given the lack of relevant studies, we undertook a qualitative synthesis.
Subgroup analysis and investigation of heterogeneity
If data had allowed, we had planned to group the results according to the type of remuneration system. However, we were unable to undertake subgroup analyses due to the lack of eligible studies.
Sensitivity analysis
In order to determine the robustness and consistency of the results, we had planned to compare RCTs (when at low risk of bias) to other studies, had we identified sufficient studies.
Results
Description of studies
Characteristics of the included studies are presented in detail in Appendix 2 and summarised in the Characteristics of included studies tables.
Results of the search
We identified 4737 studies from the literature search. Following two rounds of screening, we assessed 13 publications in detail (Figure 1).
Study flow diagram.
Included studies
Two studies (from five publications) met the inclusion criteria (Clarkson 2008; Coventry 1989), with data from 503 dental practices, representing 821 dentists and 4771 patients. Both were RCTs and were undertaken in the United Kingdom.
Excluded studies
We excluded eight studies (Blinkhorn 1996; Chalkley 2008; Fiset 2000; Holloway 1997; Mayer 2000; Mellor 1994; Mellor 1997; Rosen 1977); five on the basis that they were uncontrolled before‐after studies examining the impact on clinical activity before and after a change to the remuneration system. We excluded one ITS study on the basis of an inadequate number of time points between changes in the remuneration rate offered to primary care dentists. In addition, we excluded a cohort and an observational extension of the included Coventry 1989 study (See Characteristics of excluded studies tables).
Risk of bias in included studies
Details of the risk of bias assessment are provided in the Characteristics of included studies and summarised in Figure 2. We assessed both studies as being at an overall high risk of bias.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
We considered both the Clarkson 2008 and the Coventry 1989 trial to be of low risk with respect to the random sequence generation and allocation concealment. In the former trial, sampling, randomisation and analysis were conducted at arm’s length from the study base by a remote Health Services Research Unit at the University of Aberdeen. Randomisation was carried out using minimisation, involving four practice‐based variables: the deprivation category for the area of practice; the number of partners in the practice; the throughput of 11‐ to 13‐year‐olds; and the number of restorative sealant claims in 2002. As the dentists were recruited before randomisation occurred, this would reduce selection bias.
In the latter trial, a senior officer of the British Dental Association spun a coin to decide which one of each pair of areas should transfer to capitation and which should remain under fee‐for‐service. Again, the primary care dentists were recruited before randomisation.
Blinding
We judged the blinding of outcome assessors to represent a high potential for bias in both trials. In the Clarkson 2008 study, the primary care dentists acted as the outcome assessors after the patients had received the intervention.
In the Coventry 1989 trial, outcome assessors varied for different outcomes and for many, their measurement was undertaken by unblinded non‐participant dentists.
Incomplete outcome data
We judged both trials to be unclear with respect to incomplete outcome data. Although there does not appear to be an imbalance of missing data across the fee‐for‐service and control arms in the Clarkson 2008 trial, the authors did not provide any analysis, although the data was stated to be analysed using the intention‐to‐treat principle.
In the Coventry 1989 trial, only nine capitation practices and two fee‐for‐service practices withdrew. The number of dental practices which dropped out was therefore very low (4.9%) in the capitation areas and 1.2% in the fee‐for‐service areas. However, it is not clear how many children were examined, or if there was an imbalance across the two arms.
Selective reporting
We were not able to judge whether all the prespecified primary outcomes were reported for either trial; we assessed both trials as unclear risk of bias for this domain.
Other potential sources of bias
The Clarkson 2008 trial was judged to be unclear with respect to other potential sources of bias. The baseline characteristics and baseline outcomes of the arms were not statistically different at the practice level, probably due to the minimisation process. However, there was a statistically significant lower proportion of children having at least one sealant treatment in their second permanent molars at baseline in the fee‐for‐service only arm and the fee‐for‐service and education arms compared to the education only and the control arms. Even so, the primary analysis adjusted for a number of variables including the number of sealants placed in first permanent molars pre‐intervention, and found a statistically significant difference in favour of the dentists receiving fee‐for‐service remuneration. When baseline differences were not adjusted for, this did not reach statistical significance.
The Coventry 1989 trial was judged to represent a high risk of bias due to a lack of stratification at baseline. The baseline characteristics and baseline outcomes in a number of paired areas were unbalanced and a statistically significant difference was found in the level of reported dental caries. An analysis conducted after the trial had commenced found that the mean number of decayed missing and filled teeth in five‐ to six‐year‐olds and eight‐ to nine‐year‐olds was significantly greater in Salford compared to Doncaster and in Bromley compared to Wycombe; Salford and Bromley were both remunerated by capitation.
Effects of interventions
See: Summary of findings for the main comparison Fee‐for‐service remuneration for encouraging fissure sealant placement for second permanent molars in 12‐ to 14‐year‐olds; Summary of findings 2 Capitation remuneration compared to fee‐for‐service remuneration for encouraging routine continuing dental care of children
The two included studies were heterogenous and so we considered pooling of the data to be inappropriate. The results are presented separately in Summary of findings table 1 and Summary of findings table 2.
The primary outcomes of this review are measures of clinical activity, measures of health service utilisation, healthcare costs, and patient outcomes; the secondary outcomes are measures of non‐clinical behaviour of primary care dentists, measures of dental practice profitability/income, and measures of unintended consequences.
The Clarkson 2008 study reported that there was a statistically significant increase in clinical activity among those incentivised with a fee‐for‐service payment compared with the control arm, when the model was adjusted for the deprivation category for the area of practice, the number of partners in the practice, the throughput of 11‐ to 13‐year‐olds, and the number of restorative fissure sealants placed on first permanent molars at baseline. Using this model, the mean percentage of 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars was 9.8% higher (95% CI 1.8% to 17.8%) in the fee‐for‐service arm compared to the control arm. When left unadjusted, the difference in the mean percentage was 7.1%, which was not statistically significant (95% CI ‐1.9% to 16.1%). No further measures of health service utilisation or patient outcomes were reported.
The incremental cost‐effectiveness of the fee‐for‐service arm compared to the control arm was reported to be 0.10 i.e. for every extra GBP 1 spent there was a 0.1% increase in activity. Units were reported as "% change in outcome per £[GBP]", but no price year was provided, and the detail of the outcome measure was not specified, and can only be inferred to be sealant placement. This lack of precision represents a major flaw in the reporting of the study. Furthermore, the economic evaluation did not undertake any discounting, nor did it take into account the payments from the state (i.e. the fee‐for‐service payment), rather, it investigated the cost to practices (in terms of staff time and consumables) to avoid double‐counting, and the costs to parents.
In terms of clinical activity, the Coventry 1989 study reported that, in each of the pairs of areas, the mean number of filled teeth per 0‐ to 15‐year‐olds and the mean percentage of 0‐ to 15‐year‐olds having one or more teeth extracted tended to be lower in capitation areas while the mean percentage of 0‐ to 15‐year‐olds receiving active preventive advice tended to be higher.
Regarding health service utilisation, in each of the pairs of areas, the mean number of visits per 0‐ to 15‐year‐old tended to be lower in capitation areas compared to fee‐for‐service areas.
Regarding healthcare costs, in each of the pairs of areas, the mean expenditure in GBP per 0‐ to 15‐year‐old tended to be higher in capitation areas compared to fee‐for‐service areas. However, the authors reported that the mean expenditures should only be considered as approximations, and there is bias that places the capitation arm at a disadvantage. In addition, participating dentists in capitation areas referred significantly more children to the Community Dental Service compared to dentists in fee‐for‐service areas, and the cost of treating the children in the Community Dental Service would not have been taken into account in the economic analysis.
In terms of patient outcomes, in each of the pairs of areas, the mean number of decayed teeth per 14‐ to 15‐year‐old tended to be higher in capitation areas compared to fee‐for‐service areas, although this was only statistically significant in one of the pairs of areas. The authors reported that dentists working under capitation arrangements restored carious teeth at a later stage in the disease process than those working under fee‐for‐service arrangements, but this delay did not appear to compromise dental health.
It is important to note that not all of these comparisons between the pairs of areas were reported as statistically significant and the unit of analysis (e.g. dentists, patients, parents and administrators) was often not the same as the unit of randomisation, leading to unit‐of‐analysis error, where P values are artificially small. In addition, the baseline mean decayed/missing/filled permanent teeth (DMFT) and DMFT in the pairs of areas were unbalanced in two of the four pairs, with all the significant differences favouring the fee‐for‐service areas i.e. dental health tended to be better in the fee‐for‐service areas.
In terms of the secondary outcomes of this review, the Clarkson 2008 study did not report any relevant outcomes. However, the Coventry 1989 study reported several. Dentists under the fee‐for‐service system were more likely to introduce innovations into their dental practices compared to dentists under the capitation system (69% versus 56%, P ≤ 0.01), and reported a greater temptation to over‐prescribe using a 0 ‐ 100 visual analogue scale (31.0 versus 16.1, P ≤ 0.01), although, conversely, dentists under the capitation system felt a greater temptation to under‐prescribe (58.3 versus 37.7, P ≤ 0.01). These latter outcomes were self reported using a visual analogue scale that was not validated and they cannot be substantiated with an objective measure. Moreover, these pooled results from the four pairs of areas are problematic because the matched pairs of areas were very different from each other. Therefore, indicating where there were consistent trends in all pairs is more appropriate than testing the statistical significance of overall differences between capitation areas and fee‐for‐service areas.
Discussion
Summary of main results
In the Clarkson 2008 trial, there was a statistically significant increase in clinical activity (placement of sealants) among those incentivised with a fee‐for‐service payment compared with the control arm, when the model was adjusted.
In the Coventry 1989 trial, dentists working under capitation arrangements restored carious teeth at a later stage in the disease process than fee‐for‐service controls, and visits, fillings and extractions tended to be lower in capitation areas compared to fee‐for‐service areas, while preventive advice tended to be given more frequently.
Overall completeness and applicability of evidence
The results of the Clarkson 2008 study need to be interpreted in the context of a high risk of bias and indirectness; the primary care dentists received the fee‐for‐service remuneration in the first six months of the study, whilst the data was collected for a further twelve months after the start of the trial. It is possible that the effects of the fee‐for‐service remuneration could have been attenuated if the data had been collected at the limit of this 18‐month data period.
The clinical relevance of placing fissure sealants on thirteen‐year‐olds may also have had an impact on the decision to treat. It is good practice to place sealants on teeth as soon as it is possible to provide moisture control for the erupting tooth. Second molars erupt at twelve years of age, yet the average age at baseline was 13.2 years in the education arm and 13.3 years in both the fee‐for‐service and education arm. The data was also collected for 18 months. Again, the percentage of children with fissure sealants at the end of the study may have been attenuated by this and in high risk children, restorations may have already been indicated rather than sealants. Finally, there was imprecision as the sample size was lower than the required sample size calculated by the authors. While the adjusted mean percentage difference for 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars was 9.8% higher in the fee‐only group, the 95% CIs intervals were wide, with the lower boundary showing an increase of 1.8%. The clinical significance and cost‐effectiveness of a financial payment that results in an increase of 1.8% in the number of children receiving fissure sealants is difficult to determine. In addition, given that the study was conducted in the four most deprived areas of Scotland the applicability of the findings to other settings may be limited.
In the Coventry 1989 trial, the researchers did not stratify the participants, based on disease experience at baseline and there were significantly different disease levels across a number of the paired arms, in addition to unit of analysis error. This means that it was not possible to determine the impact that the remuneration systems had on the health of the children, nor determine the cost‐effectiveness of either arm, given the unknown impact on patient outcomes.
Another limitation of the review is the inclusion criteria regarding study design. Major advances in econometrics have been made, which make it possible to draw causal inferences from non‐random assignments of patients and dentists, for example, as demonstrated by Chalkley 2008. These studies could be considered and triangulated with experimental evidence to fully inform the evidence base. Future updates of this review may also consider broadening the inclusion criteria to consider data from non‐experimental sources.
Quality of the evidence
The number of studies using an experimental design was very low. Both included studies had a high risk of bias and the quality of the evidence from the two included studies was low/very low for all outcomes, as assessed by GRADE.
Potential biases in the review process
Bias in the review process was kept to a minimum. Three authors (PRB, AMG and JP) screened the titles and determined inclusion, assessed for bias and extracted the data. Any differences were resolved by discussion. No post hoc changes were made to the review methods described in the protocol.
One area which may introduce bias is the choice of grey literature sources. The identified sources do not represent a comprehensive list of international health economics centres and is unclear as to whether this would have introduced some form of reporting bias within the review. Future updates of the review will aim to identify further relevant sources of both published and unpublished papers.
Agreements and disagreements with other studies or reviews
An examination of the health economic literature would suggest that a retrospective cost‐based system like fee‐for‐service shifts the cost of care to the third party payer and creates an incentive for over‐provision of services, as activity generates revenue. In contrast, prospective payment systems are said to cut the link between the revenue per case and create an incentive for under‐provision, with a restriction of services largely to those with low needs (cream‐skimming), the "dumping" of high cost patients and the "skimping" or under‐provision for those with high needs (Ellis 1997; Krasnik 1990).
In Gosden et al's review, fee‐for‐service payments were associated with an increase in the quantity of primary care services, but changes to patient outcomes were equivocal and there was considerable variation in the study setting and the range of outcome measures utilised (Gosden 2000). In Scott et al's review, there was insufficient evidence to determine the impact of financial incentives on the quality of primary health care and the quality of the included studies was considered poor (Scott 2011). Flodgren et al's overview of reviews concluded that financial incentives may be effective in changing healthcare professional practice, but the included studies were of low to moderate quality and there were no studies evaluating patient outcomes (Flodgren 2011).
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
| Population: Dentists seeing children with erupted second permanent molars Control: No specific intervention | |||||
| Outcomes | Illustrative comparative risks (95% CI) | Relative effect | No. of Participants | Quality of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Control | Fee‐for‐service remuneration | ||||
| Measures of clinical activity: mean percentage of 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars per dentist (weighted by number of children seen) ‐ adjusted1 | ‐ | RD 9.8% higher (1.8% higher to 17.8% higher)3 | ‐ | 133 dentists | ⊕⊝⊝⊝ |
| Measures of clinical activity: mean percentage of 12‐ to 14‐year‐olds receiving fissure sealants for second permanent molars per dentist (weighted by number of children seen) ‐ unadjusted2 | 26.3% (CI NR) | RD 7.1% higher (1.9% lower to 16.1% higher) | RR 0.27 (CI NR) | 133 dentists | ⊕⊝⊝⊝ very low4 |
| Healthcare costs: cost‐effectiveness of fee‐for‐service vs. control (reported as the "% change in outcome per £[GBP]" ‐ currency year NR)5 | ‐ | ‐ | 0.10 (CI NR) | 68 dentists | ⊕⊝⊝⊝ very low4 |
| CI: Confidence interval; GBP: Pound Sterling; NR: Not reported; RD: Risk difference; RR: Risk ratio | |||||
| GRADE Working Group grades of evidence | |||||
| 1The model adjusted for the baseline dental practice‐level covariates (deprivation category for the area of dental practice, number of partners in practice, throughput of 11‐ to 13‐year‐olds and the number of restorative fissure sealants placed on first permanent molars at baseline). 2The basis for the assumed risk is the risk in the control group (i.e. the probability of a dentist in the control group fissure sealing a second permanent molar of a 12‐ to 14‐year‐old). The corresponding risk (the risk difference) is based on the assumed risk in the control group and the relative effect of the fee‐for‐service remuneration (the risk ratio). 3Statistically significant at the 5% level.
5Outcome in the incremental cost‐effectiveness ratio (ICER) is not specified and can only be assumed to relate to sealant placement | |||||
| Population: Children undergoing routine continuing dental care at 354 dental practices (the number of dentists varied over time as dentists joined and left dental practices, so the number of dental practices was the stable and primary parameter; the total number of 0‐ to 15‐year‐old children was not reported accurately) | |||||
| Outcomes | Illustrative comparative risks (95% CI) | Relative effect | No. of participants | Quality of the evidence | |
| Assumed risk | Corresponding risk | ||||
| Fee‐for‐service remuneration | Capitation remuneration | ||||
| Measures of health service utilisation: mean number of visits per 0‐ to 15‐year‐old | Northern urban community: 2.8 | MD 0.4 lower1 | ‐ | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 2.3 | MD 0.4 lower1 | ||||
| Rural community: 2.5 | MD 0.3 lower | ||||
| Scottish community: 2.8 | MD 0.5 lower1 | ||||
| Patient outcomes: mean number of filled teeth per 0‐ to 15‐year‐old | Northern urban community: 0.78 | MD 0.18 lower1 | ‐ | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 0.34 | MD 0.03 lower | ||||
| Rural community: 0.44 | MD 0.21 lower | ||||
| Scottish community: 0.91 | MD 0.28 lower1 | ||||
| Patient outcomes: mean percentage of 0‐ to 15‐year‐olds having one or more teeth extracted3 | Northern urban community: 18% | RD 5% lower1 | RR 28% lower1 | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 7% | RD 0% | RR 0% | |||
| Rural community: 10% | RD 3% lower1 | RR 30% lower1 | |||
| Scottish community: 15% | RD1% lower | RR 7% lower | |||
| Patient outcomes: mean number of decayed teeth per 14‐ to 15‐year‐old (data for 0‐ to 15‐year‐olds NR) | Northern urban community: 0.16 | MD 0.16 higher | ‐ | 1919 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 0.24 | MD 0.07 higher | ||||
| Rural community: 0.58 | MD 0.75 higher2 | ||||
| Scottish community: 0.65 | MD 0.15 higher | ||||
| Measures of clinical activity: mean percentage of 0‐ to 15‐year‐olds receiving active preventive advice3 | Northern urban community: 19% | RD 27% higher1 | RR 142% higher1 | ˜ 2250 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 18% | RD 15% higher | RR 83% higher | |||
| Rural community: 34% | RD 5% lower | RR 15% lower | |||
| Scottish community: 28% | RD 9% higher | RR 32% higher | |||
| Healthcare costs: mean expenditure in GBP (currency year NR) per 0‐ to 15‐year‐old5 | Northern urban community: 20.55 | MD 4.22 higher | 21% higher | 276,4145 (1 study4) | ⊕⊕⊝⊝ low6 |
| Commuter suburb community: 16.67 | MD 6.18 higher | 37% higher | |||
| Rural community: 17.29 | MD 6.90 higher | 40% higher | |||
| Scottish community: 17.68 | MD 1.52 higher | 9% higher | |||
| CI: Confidence interval; GBP: Pound Sterling; MD: Mean difference; NR: Not reported; RD: Risk difference; RR: Risk ratio | |||||
| GRADE Working Group grades of evidence | |||||
| 1Statistically significant at the 5% level. However, the unit of analysis (e.g. dentists, patients, parents and administrators) was often not the same as the unit of randomisation. This leads to unit‐of‐analysis error, where P values are artificially small (though the estimates of effect are unbiased), leading to false positive conclusions that the intervention had an effect. 2Statistically significant at the 1% level. However, the unit of analysis (e.g. dentists, patients, parents and administrators) was often not the same as the unit of randomisation. This leads to unit‐of‐analysis error, where P values are artificially small (though the estimates of effect are unbiased), leading to false positive conclusions that the intervention had an effect. 3The basis for the assumed risk is the risk in the control group (i.e. the probability of a dentist in the control group giving preventive advice to or extracting a tooth for a 0‐ to 15‐year‐old). The corresponding risk (the risk difference) is based on the assumed risk in the control group and the relative effect of the capitation remuneration (the risk ratio). 4There were four matched pair of Health Service administrative areas. These randomised pairs were treated as separate, thus the overall study contained four replicates under contrasting socioeconomic and environmental circumstances. There were only two Health Service administrative areas randomised in each replicate, therefore each arm of each replicate only contained one Health Service administrative area. 5 All payments made to study dentists for the treatment of 0‐ to 5‐year‐olds during 1988 were divided by the estimated numbers of children treated. However, the estimated number is only an approximation as it was impossible to eliminate double‐counting, particularly in the fee‐for‐service system. This means that the mean expenditure per 0‐ to 15‐year‐old should only be considered as close approximations, and there is bias that places the capitation arm at a disadvantage. In addition, participating dentists in capitation areas referred significantly more children to the Community Dental Service compared to dentists in fee‐for‐service areas; this is despite the fact that non‐participating dentists in the capitation areas tended to refer significantly fewer children compared to non‐participating dentists in fee‐for‐service areas. The cost of treating the children in the Community Dental Service would not have been taken into account in the economic analysis. 6 Quality of the evidence
| |||||
