Effect of screening for type 2 diabetes on healthcare costs: a register-based study among 139,065 individuals diagnosed with diabetes in Denmark between 2001 and 2009

Abstract

Background

Trials have not demonstrated benefits to the population of screening for type 2 diabetes. However, there may be cost savings for those found to have diabetes.

Aim

To compare healthcare costs among incident cases of type 2 diabetes in a screened group with those in an unscreened group.

Methods

In this register-based non-randomised controlled trial, eligible individuals were all men and women aged 40-69 years without known diabetes, registered with a general practice in Denmark (n=1,912,392). Between 2001 and 2006, 153,107 individuals registered with 181 practices participating in the Anglo–Danish–Dutch Study of Intensive Treatment in People with Screen-Detected Diabetes in Primary Care (ADDITION)-Denmark study, were sent a diabetes-risk-score questionnaire. Individuals at moderate-to-high risk were invited to visit their family doctor for assessment of diabetes status and cardiovascular risk (screening group). The 1,759,285 individuals registered with all other practices in Denmark constituted the retrospectively constructed no-screening (control) group. In this post hoc analysis, we identified individuals from the screening and no-screening groups who were diagnosed with diabetes between 2001 and 2009 (n=139,065). Using national registry data, we quantified the cost of health care services in these two groups between 2001 and 2012. From a health care sector perspective we estimated potential health care cost savings for individuals with diabetes attributable to the screening programme.

Results

In the screening group, 27,177 of 153,107 (18% of those sent a risk score questionnaire) individuals attended for screening, of whom 1,533 were diagnosed with diabetes. Between 2001 and 2009, 13,992 people were newly diagnosed with diabetes in the screening group (including those diagnosed by screening) and 125,073 in the no-screening group. Health care costs were significantly lower in the screening group compared to the no-screening group (difference in mean total annual health care costs: -889 € per patient with incident diabetes, (95% CI -1,196: -581). The screening programme was associated with a cost saving per patient with incident diabetes over a five-year period of -2,688 € (95% CI -3,995; -1,421).

Conclusion

Healthcare costs were lower among incident cases of type 2 diabetes in a screened group compared to an unscreened group. The relatively modest cost per discovered diabetes patient was offset within two years by savings in the healthcare system.

Introduction

While modelling studies indicate that screening for type 2 diabetes might be effective and cost-effective [1–3], trials of population-based screening for type 2 diabetes [4] and related cardiovascular risk factors [5] have not demonstrated beneficial effects at the population level. However, there appear to be benefits for those found to have diabetes. We have previously shown that screening for type 2 diabetes and cardiovascular risk factors does not reduce mortality and cardiovascular disease in the general population in Denmark [5], however for individuals diagnosed with diabetes, screening was associated with a reduction in mortality and cardiovascular disease risk [6]. Using data from Danish national registers, and a retrospectively constructed control group, we extended this analysis to compare total healthcare costs for individuals with incident diabetes in the screening and no-screening groups and estimated the potential cost savings of the ADDITION intervention from a healthcare sector perspective.

Methods

ADDITION-Europe is a cluster-randomised trial comparing the effects of screening for type 2 diabetes followed by intensive treatment or routine care [7, 8]. We report results from a post hoc analysis using data from the screening phase of the Danish arm of the study in conjunction with data from Danish national registers. Ethical approval for the ADDITION-Denmark study was granted by a local scientific committee (number 20000183). As this was a registry-based study using anonymised data, participants did not give informed consent. This approach was approved by the Danish Data Protection Agency and the Danish Health and Medicine Authority.

Screening programme

Full details of the programme have been reported elsewhere [5, 6, 9]. In brief, we performed a population-based stepwise screening programme in people aged 40 to 69 years, without known diabetes between 2001 and 2006 [7, 8, 10]. All general practices, (GPs), in five out of 16 counties in Denmark (Copenhagen, Aarhus, Ringkoebing, Ribe and South Jutland) were invited to take part in ADDITION-Denmark (n=744); 209 (28.1%) accepted.

Eligible individuals registered with the 181 practices, that agreed to take part, were sent a diabetes risk-score questionnaire [8, 10], with an invitation to visit their general practitioner (GP) for a diabetes test and a cardiovascular risk assessment if they scored ≥ 5 (maximum 15 points), or were invited when visiting the practice for another reason (n=35 practices). No reminders were sent. Participants who attended a screening appointment underwent measurement of height, weight, and blood pressure. A capillary blood sample was taken for testing of random blood glucose (RBG). A venous blood sample was taken for measurement of total cholesterol and HbA_1c. GPs were encouraged to calculate the European Heart SCORE [11] during the appointment, to inform patients about their score and provide appropriate advice and treatment to those at high risk. Individuals with an RBG ≥5.5mmol/l or HbA_1c ≥ 5.8% were invited to return to the practice for a fasting blood glucose (FBG) capillary test. An OGTT was performed at the same consultation if FBG was 5.6–6.1 mmol/l and/or HbA_1c≥5.8%. WHO 1999 criteria were used to diagnose diabetes[12].

In the intervention group, participants diagnosed with type 2 diabetes were subsequently managed according to the treatment regimens to which their practice was allocated: routine care or intensive treatment[9]. Intensive treatment included small group or practice-based educational meetings with GPs and nurses to discuss treatment targets, algorithms and lifestyle advice; audit and feedback in group meetings up to twice per year or coordinated by post; and educational materials for patients. GPs were trained in motivational interviewing to encourage lifestyle change in patients in a 1.5 day educational seminar.

Sampling frame

We identified all eligible individuals in the original ADDITION-Denmark study (n=153,107), including those who did not attend for screening, in the Danish National Diabetes Register [13] (the screening group). Using the same registry, we also identified all individuals aged 40-69 years without known diabetes who, between 2001 and 2006, were registered with general practices that were not invited to take part in ADDITION-Denmark or who declined to take part in ADDITION-Denmark (n=1,759,285) (the no-screening control group). We then identified individuals from the intervention and no-intervention groups who were diagnosed with incident diabetes between 1 January 2001 and 31 December 2009 (Figure One). We included individuals diagnosed with diabetes during this period based on recent estimates of lead time, which suggest that there is around 2.2 years between detection by screening and clinical diagnosis [6]. Including individuals diagnosed with diabetes in the 3 years (2006-09) following the end of the ADDITION screening phase (2001-2006) would therefore capture most individuals in the no-intervention group who could have been diagnosed by screening if they had been in the screening group. Our definition of incident diabetes for both groups was a proxy measure based on date of inclusion in the Danish National Diabetes Register [14].

Figure One.

Visual representation of sampling frame. The “S” in a blue circle denotes individuals detected by the ADDITION stepwise screening programme. The “C” in a red circle denotes clinically-diagnosed diabetes cases.

Data

We linked information about individuals diagnosed with diabetes to other Danish registers using unique civil registration numbers. We retrieved information on age, sex, education, immigration/emigration, citizenship, redeemed cardioprotective medication and chronic disease. Education was categorised according to Unesco’s International Standard Classification of Education [15]. We grouped data on citizenship into European and non-European as a proxy for ethnicity. Data on health care service usage in primary and secondary care, as well as information on redeemed medication, were retrieved from the National Patient register, the National Health Insurance Register and the National Prescription Register [16].

Costs

The costs of ADDITION-Denmark include the cost of the screening programme and the cost of the intensive treatment programme (for those found with screen-detected diabetes and randomised to this arm). Health care costs include total costs of health care service usage and redeemed medication for all individuals in our sampling frame during the follow-up period. All costs were calculated in Danish krone (DKK) and converted to Euro (conversion rate 1€=7.44 DKK) from a health care sector perspective e.g. all costs included overhead costs and patients’ own payment for redeemed medication.

The cost of the screening programme has previously been reported [17]. The average cost of detecting an individual with previously undiagnosed type 2 diabetes was estimated to be 936€ [17] including the cost of sending the invitation letter, the cost of the screening consultations and the laboratory tests. We multiplied GP fees by 1.438 to capture overhead costs as Danish GPs are paid on a combined capitation and fee for service scheme. The cost of the intensive treatment intervention includes the cost of small group or practice-based educational meetings with GPs and nurses to discuss treatment targets, algorithms and lifestyle advice; audit and feedback in group meetings up to twice per year or coordinated by post; educational materials for patients; and small financial incentives for GPs [9]. In addition, 64 GPs participated in a 1.5 days educational seminar on motivational interviewing. Summing the costs of the screening programme and the intensive treatment costs provides an estimate of 967€ per patient with incident diabetes. The cost of healthcare usage was defined as opportunity costs using market values (excluding VAT) [18, 19]. We quantified the health care costs in the screening and no-screening groups using individual-level register data. These data allowed us to estimate the cost of (i) inpatient health care utilisation (data available for 2007-2012); (ii) out-patient health care utilisation (data available for 2007-2012); (iii) primary care utilisation (data available for 2001-2012), and (iv) redeemed medications (data available for 2001-2012). Table One describes the different cost units and our methods for calculating them.

Table One. Cost units and methods of calculation for cost components.

Cost component	Cost unit and method of calculation
Inpatient and outpatient services delivered in Danish hospitals registered in the national hospital register	Diagnosis Related Grouping system (DRG) tariffs and Danish Ambulatory Grouping System (DAGS) were applied[28]. In the study costs are understood as opportunity costs i.e. resources are scarce and resources spent in one area cannot be spent in another [18]. Activities in the Danish health care sector have been converted into a monetary cost through the use of the DRG. The DRG-tariffs are calculated from average operating costs within each DRG-group on a nationwide level [28]. The system is developed for settlement of accounts for patients in hospitals, patients who have chosen a hospital in another region than their region of residence and for patients in private hospitals. This calculation produces approximate average costs, which should be used in economic evaluations with caution [29].
Primary care services delivered by general practitioners and privately practicing specialists. Visits by GP, physiotherapists, chiropractors, chiropodists, opticians, and dieticians.	Reimbursement fees between the National Health Insurance scheme and private practicing physicians are used as cost units. General Practitioners are compensated by regions through a combination of per capita fee (app.30% of total) and fee for service (app. 70%) [30]. To reflect this payment scheme in the unit cost, 43.8% of the fee for service in general practice was added to the cost. Overhead costs covered by the capitation fee were not distributed across the numbers of visits but by resource burden.
Prescribed pharmaceuticals dispensed by Danish pharmacies and registered in DNPrR. (Pharmaceuticals consumed in hospitals are included in DRG-tariffs. Over-the-counter drugs are not included in this analysis).	Total Sales Price (TSP) includes patient Out of Pocket Payments (OPP). Costs of prescribed pharmaceuticals are shared between the patient and the primary health care sector by a copayment scheme in Denmark where patients are reimbursed according to their need. These costs were aggregated since total costs are measured regardless of who pays. While pharmaceutical consumption includes value added tax (VAT), Danish health care services are exempted from VAT. Given that VAT is 25% in Denmark, 20% of pharmaceutical consumption was subtracted to calculate comparable net costs.

Statistical analysis

We summarised characteristics of all patients diagnosed with incident diabetes between 2001 and 2009 separately in the screening and no-screening groups. Date of entry to the study was the date of inclusion in the diabetes register. All analyses were completed using Stata Version 14.1 (STATA Corp., College Station, Texas, USA). Statistical significance was inferred at a two-tailed p<0·05. We calculated robust standard errors to allow for clustering of patients within practices.

Cost analysis

We quantified healthcare usage and calculated average healthcare costs in each group over the follow-up period by dividing total healthcare costs for each individual by the time they spent in the study. As data from the health registers was available for different time periods, we conducted analyses for two time periods: 2001-2012 (with an average of 7.0 years follow-up, N = 139,048) and 2007-2012 (with an average of 5.2 years follow-up, N = 133,307). In this second analysis, we only included individuals who were alive and living in Denmark on 1 January 2007. To examine potential differences in healthcare costs between the groups, we used ordinary least squares regression (OLS), adjusting for age, sex, education and GP county. We also controlled for prevalent chronic disease at diagnosis (IHD, stroke, cancer). We quantified costs from more than three years before diagnosis to more than six years after diagnosis using difference in difference regression on panel data (data covering more years over a time period) with mean correction for confounders. The difference-in-difference method allowed us to estimate the “isolated” effect of ADDITION intervention on healthcare costs, where we examine the differences between the screening population and no-screening population within the periods before and after diagnosis. We included mean-adjusted covariates in the regression, where we applied effect coding for the categorical variables [20] in order to adjust for the demographic, epidemiological and regional differences. Therefore, we were able to interpret the constant term as the predicted healthcare costs for an individual in the no-screening group in the year of diagnosis.

To estimate the potential economic benefits of the screening programme among individuals with incident diabetes we subtracted the cost of the screening program from the estimated health care cost savings accruing in the screening group. We calculated cost savings over a five-year period following the screening phase using an annual discount rate of 5%.

Sensitivity analysis

We also examined the difference in healthcare costs between the screening population assigned to routine care only and the no-screening population.

Results

Population characteristics

Of 153,107 eligible people in the screening group who were sent a diabetes risk score questionnaire, 27,177 (18%) attended their GP for a diabetes test and a cardiovascular risk assessment. 1,533 participants (1% of those eligible for screening) were diagnosed with diabetes; 1,406 of these were subsequently included in the diabetes register. There were 1,759,285 individuals in the no-screening group. Between 1 January 2001 and 31 December 2009, 139,065 people from our sampling frame were diagnosed with incident diabetes and included in the Danish National Diabetes Register. Of these, 13,992 (10.1%) were in the screening group and 125,073 (89.9%) in the no-screening group. The groups were well balanced for age and citizenship (Table Two). There were slightly less men in the screening group (53.6%) compared to the no-screening group (56.4%). A larger proportion of the screening group had received 15+ years of education. Slightly lower proportions of the screening group had experienced IHD, stroke or cancer before diabetes diagnosis compared to the no-screening group.

Table Two. Characteristics of individuals with diabetes by intervention group.

	Screening group n=13,992	No-screening (control) group n=125,073
Mean age at diagnosis (SD), years	59.9 (7.7)	59.2 (9.2)
Male sex, n (%)	7,495 (53.6)	70,552 (56.4)
Years of education, n (%)
0 to 10	5,610 (40.1)	57,765 (44.6)
10 to 15	6,237 (44.6)	55,226 (44.2)
15+	2,145 (15.3)	14,082 (11.3)
European citizenship, n (%)	13,944 (99.7)	123,849 (99)
Previous IHD, n (%)1	1,586 (11.3)	16,216 (13)
Previous stroke, n (%)1	628 (4.5)	6,851 (5.5)
Previous cancer, n (%)1	2,027 (14.5)	19,276 (15.4)

¹Data taken from the National Patient Registry; data included from 1994 until date of diabetes diagnosis; IHD = ischaemic heart disease

Healthcare costs in the screening and no-screening group

Figures Two, Three, Four and Five show the mean adjusted annual costs per patient with incident diabetes by health care sector both prior to and following the time of diagnosis in the screening and no-screening groups. The results from the difference in difference regression analyses are presented in Supplementary Table one.

Figure Two.

Pharmaceutical costs per patient with incident diabetes 2001-2012 by intervention group according to time from diagnosis*.

* Dashed line indicates costs of no-screening (control) group; solid line represents costs of screening group. Black circles indicate that estimated mean differences in costs between no-screening and screening groups in each time period are significantly different from zero at 5% level; white circles indicate no statistically significant estimated mean differences in costs in each time period at >5% level.

Figure Three.

Primary care costs per patient with incident diabetes 2001-2012 by intervention group according to time from diagnosis*.

* Dashed line indicates costs of no-screening (control) group; solid line represents costs of screening group. Black circles indicate that estimated mean differences in costs between no-screening and screening groups in each time period are significantly different from zero at 5% level; white circles indicate no statistically significant estimated mean differences in costs in each time period at >5% level.

Figure Four.

Inpatient care costs per patient with incident diabetes 2007-2012 by intervention group according to time from diagnosis.*

* Dashed line indicates costs of no-screening (control) group; solid line represents costs of screening group. Black circles indicate that estimated mean differences in costs between no-screening and screening groups in each time period are significantly different from zero at 5% level; white circles indicate no statistically significant estimated mean differences in costs in each time period at >5% level

Figure Five.

Outpatient care costs per patient with incident diabetes 2007-2012 by intervention group according to time from diagnosis*.

* Dashed line indicates costs of no-screening (control) group; solid line represents costs of screening group. Black circles indicate that estimated mean differences in costs between no-screening and screening groups in each time period are significantly different from zero at 5% level; white circles indicate no statistically significant estimated mean differences in costs in each time period at >5% level.

Following diagnosis, healthcare usage and subsequent costs were significantly higher in the no-screening group compared to the screening group. For example, primary care costs were approximately 5% higher in the no-screening compared to the screening group during follow-up. The costs of redeemed medication were already higher in the no-screening group before diagnosis but the difference between the groups increased further following diabetes diagnosis. Costs were approximately 12% higher in the no-screening group compared to the screening group in the six years following diagnosis. For in-patient and out-patient visits there were no clear trends associated with the point of diabetes diagnosis. However, on average, costs were lower among the screening group compared to no-screening group both prior to and following diagnosis.

The mean annual cost estimates for primary care and redeemed medication were based on 2001-2012 data, while inpatient and outpatient costs were based on 2007-2012 data. The mean total cost difference is therefore estimated as the sum of the four mean annual cost components: in- and outpatient cost, primary care cost and cost of redeemed medication. After adjustment for confounders, annual healthcare costs per patient with incident diabetes were significantly lower in the screening compared to the no-screening group across all healthcare sectors and for redeemed medication: annual difference in inpatient care costs -662 €, (95% CI -865;-459), outpatient care costs -82 € (95%CI -157;-7), primary care costs -51€, (95%CI -63; -38) and pharmaceuticals -94 € (95%CI -111; -76). The difference in mean total annual health care costs per patient with incident diabetes between groups was -889 €, (95% CI -1,196; -581).

After subtracting the cost of introducing the ADDITION screening programme (967 € per patient with incident diabetes) from the estimated annual health care cost savings (889 € per patient with incident diabetes), and discounting this figure in the five years following the intervention, there was a cost saving associated with each of the patients with incident diabetes over a five-year period of -2,688 € (95% CI -3,995; -1,421). This finding suggests that the intervention was associated with cost savings in the health care system within two years of the screening programme being introduced. Based on the 13,992 individuals with diabetes in the screening group, we estimate that the intervention saved the Danish health care system almost 37.6 million €, (95% CI 19.8; 55.3 million) in the five years following the introduction of the screening programme.

Sensitivity analysis

In our sensitivity analysis, the annual healthcare costs of the screening population assigned to routine care remained lower than costs in the no-screening group for: inpatient care -633 €, (95% CI -898; -368), primary care -41 €, (95% CI -58; -24), and medication -90 € (95% CI -117; -63). The outpatient care costs were similar in both groups. As the costs of the screening programme within the routine care group were 936€ per individual, and total mean annual difference in costs among screening and no screening groups was -752 € (96% CI -1,169; -335) per individual, ADDITION brings cost savings of 2,158 € (CI 444; 3,871) per individual within 5 years following the screening period. For the whole population of 13,992 individuals, the cost savings would be 30.1 million € (CI 6,2; 54,1 million).

Discussion

Our cost analysis of the ADDITION-Denmark screening programme showed that health care costs were significantly lower for individuals with incident type 2 diabetes in the screening group compared to the no-screening group. This was true for all examined health services - inpatient, outpatient, and primary care - as well as for medication costs. We also showed that within two years of being introduced, ADDITION-Denmark was associated with cost savings in the health care system.

Our estimates of annual healthcare costs for individuals with diabetes in Denmark align well with published literature. For example, a previous study [21] showed that the cost per diabetes patient per year was 6,390 € in 2011 (compared to an annual mean of 5,910 € for the screening group and 6,854 € for the no-screening group between 2007-2012).

In ADDITION-Denmark the relatively modest cost per discovered diabetes patient was offset within two years by savings in the healthcare system. In terms of comparison, there are no other published trials that examine the costs of healthcare among incident cases of type 2 diabetes in a screened group with those in an unscreened group. However, modelling studies show that screening for type 2 diabetes at the population level would be cost-effective over 30 and 50 years [1, 2, 22].

Herman et al argue that that the benefits of screening and treatment primarily accrue from early diagnosis and by hastening the treatment of CVD risk factors in the lead time [3]. We have previously reported that ADDITION-Denmark brought forward the diagnosis of diabetes by 2.1 years and was associated with a significant reduction in mortality and CVD among those with diabetes [5]. We observed an increase in the proportion of screen-detected individuals who redeemed cardio-protective medication during follow-up [6]. However, larger proportions of clinically diagnosed individuals in the no-screening group redeemed medication compared with clinically diagnosed individuals in the screening group. As individuals in the no-screening group were diagnosed at a later stage in the disease trajectory, they may have had higher cholesterol, blood glucose and blood pressure values at diagnosis compared with the screening group, necessitating higher levels (and cost) of cardioprotective medication. The promotion of healthy behaviour change might also have impacted on CVD risk factors and CVD and mortality rates and consequently lower costs of healthcare among diabetes patients in the screening group.

New drugs are more expensive which might increase the cost savings based on these study data. However, new agents tend to be restricted to second and third line and hence may not have a major impact on our estimates of cost-savings associated with screening in the first 5-10 years of the disease trajectory. In addition, preventive strategies for diabetes have gained traction in the past decade and may be associated with further cost savings. For example the Diabetes Prevention Program in the U.S [22], is being implemented through a number of different delivery channels [23, 24]. NHS England also launched a National Diabetes Prevention Programme in 2016, which now covers 75% of the population [25, 26]. Modelling studies suggest that these programmes will be cost-effective in the long-term.

Strengths and limitations

In this study, we used real-world, individual patient-level data to examine healthcare costs among incident cases of type 2 diabetes in a screened group compared to an unscreened group. We applied a healthcare sector perspective in our cost analysis. Only costs related to health care services and redeemed medication were included in the study. As such, costs relating to patients’ lost labour market productivity, utilisation of nursing services or other services such as transportation and helping aids in the house were not included in our analysis. These costs have previously been demonstrated to be high among diabetes patients, making up as much as 80% of the total societal costs [21, 23]. Furthermore, we did not include quality of life or the value of higher life expectancy in our analyses. Data on quality of life would enable better comparison to the literature on the cost utility of screening and prevention [22, 24–27].

We did not include the health care cost savings accruing from an individuals’ death, as it will always be less expensive if a patient dies and hence there is no argument for treatment or prevention [18].

A limitation of our study was the non-randomised design; we cannot eliminate the possibility of selection bias and residual confounding. Groups were well balanced for the characteristics investigated at baseline. However, our findings might have been influenced by the higher levels of education and the slightly lower levels of pre-existing chronic disease in the intervention group. The cost of redeemed medication was also lower in the screening group before the introduction of the ADDITION-Denmark screening intervention. We did adjust for age, sex, education, GP county and prevalent chronic disease at diagnosis, which impacted on the cost difference size, although the final result remained significant. It is likely that adjusting for county took account of some of the potential socio-economic differences across different regions in Denmark. We tried to minimize lead and length time biases by comparing costs for all individuals diagnosed with diabetes in the screening and no-screening groups.

We could have applied bootstrapping and cost acceptability curves to handle uncertainty [18]. As the intervention proved to be cost saving within a short time period we chose to use the lowest levels of confidence intervals to describe uncertainty instead.

Conclusion

Healthcare costs were lower among incident cases of type 2 diabetes in a screened group compared to an unscreened group. The relatively modest cost per discovered diabetes patient was offset within two years by savings in the healthcare system.

Research in context.

What is already known about this subject?

• Trials have not demonstrated benefits to the population of screening for type 2 diabetes.

• However, for individuals diagnosed with diabetes, screening is associated with a reduction in mortality and cardiovascular disease risk. The potential effect on healthcare costs in this group is unknown.

What is the key question?

• What is the effect of screening on healthcare costs among individuals diagnosed with diabetes in Denmark?

What are the new findings?

• Healthcare costs were lower among incident cases of type 2 diabetes in a screened group compared to an unscreened group.

How might this impact on clinical practice in the foreseeable future?

• Our findings add to the evidence base on the potential cost savings of early detection among individuals with incident type 2 diabetes.

Abbreviations

ADDITION

Anglo–Danish–Dutch Study of Intensive Treatment in People with Screen-Detected Diabetes in Primary Care

CI

Confidence Interval

CVD

Cardiovascular Disease

FBG

Fasting Blood Glucose

GP

General Practice/General Practitioner

HbA_1c

Haemoglobin A1c/glycated haemoglobin

IHD

Ischemic Heart Disease

RBG

Random Blood Glucose

WHO

World Health Organization