Cost of microvascular complications in people with diabetes from a public healthcare perspective: a retrospective database study in Brazil

Abstract

Aim

To evaluate direct medical costs incurred by patients with diabetes in the periods before and after experiencing a microvascular complication from a Brazilian public healthcare system perspective.

Materials and methods

This was a retrospective, observational study using the Brazilian Unified Health System (DATASUS) database. Direct medical costs (hospitalization and outpatient) were extracted for patients with evidence of diabetes and a microvascular complication (January 2012–December 2018) and converted to 2019 US Dollars (USD). Length of hospital stays was also extracted. Mixed-effects logistic regression explored associations between demographic/clinical characteristics and incurrence of high direct medical costs (defined as the highest tertile of the annual costs ranked by median cost in the total population).

Results

In total, 2,096 patients with diabetes experienced a microvascular complication and met study inclusion/exclusion criteria. Median [interquartile range] annual costs (USD/patient) were 176.3 [91.0; 481.2] at baseline, increasing to 1,678.5 [287.0; 6,908.4] and 5,172.4 [274.8; 7,395.9] in the first and second year after the complication, respectively. Median hospital stay was 2.0 and 3.0 days at baseline and in the first year, respectively. The odds of incurring high costs were substantially elevated in the first and second years (odds ratios of 69.9 and 84.7, respectively, vs. baseline, both p < .001).

Limitations

The DATASUS database covers secondary and tertiary care (not primary), adding selection bias to our sample. Additionally, our findings may not apply to the entire Brazilian population, as around 25% have some access to private healthcare.

Conclusions

This study demonstrates a large increase in costs, from the perspective of the Brazilian public healthcare system, in patients with diabetes after experiencing a microvascular complication compared with pre-complication costs. In addition to providing up-to-date cost estimates, our findings highlight the need to appraise the cost-effectiveness of evidence-based strategies that reduce the risk of diabetes-related microvascular complications in Brazilian patients.

Keywords:

• Diabetes mellitus
• microvascular complications
• cost and cost analysis
• healthcare expenditure
• healthcare utilization
• Brazil

JEL CLASSIFICATION CODES:

• C25
• C2
• C
• I10
• I1
• I

Introduction

The global prevalence of diabetes has steadily increased over recent decades to an estimated 463 million people in 2019 and is predicted to reach 700 million by 2045^1,2. It is now considered one of the most prevalent and costly non-communicable diseases worldwide, imposing a substantial societal burden due to high medical costs, lost productivity, impaired quality of life, and premature death^3–6. Low- and middle-income countries are disproportionally impacted by diabetes. It is estimated that four of every five individuals with diabetes currently live in a low- or middle-income country, and the prevalence of diabetes is increasing fastest in such countries^1,7. Brazil is a large, upper-middle-income country and was ranked fifth globally in 2019 for the number of people with diabetes¹. Around 16.8 million people were living with diabetes in Brazil in 2019, equivalent to an age-adjusted prevalence of 10.4% among those aged between 20 and 79 years¹. In terms of its economic burden, Brazil has the third most expensive diabetes-related healthcare bill globally, estimated to be 52.3 billion United States dollars (USD) in 2019¹.

Diabetes is associated with debilitating, long-term complications that account for a high proportion of medical costs and health burden^5,8. For example, in the United Kingdom, the treatment and management of diabetes-related complications account for 78.9% of disease-specific direct medical costs⁹. Both macrovascular (including chronic heart disease, peripheral artery disease, and stroke) and microvascular (including nephropathy, neuropathy, and retinopathy) complications of diabetes are common and associated with significant morbidity, mortality, and healthcare costs^8,10–12. Cardiovascular (CV) disease remains the largest cause of disability and mortality among patients with diabetes¹³. Microvascular complications account for significant morbidity and considerably impair quality of life¹⁰.

Evidence-based diabetes treatment guidelines now address the co-management of type 2 diabetes and microvascular complications^14,15. There is robust evidence that appropriate glycemic control can reduce or delay the appearance of microvascular complications, thereby decreasing the clinical burden and associated costs^14,16. Additionally, in patients with type 2 diabetes and chronic kidney disease (CKD), guidelines advise the use of a sodium–glucose cotransporter-2 inhibitor or glucagon-like peptide-1 receptor agonist, as these have been shown to reduce the risk of CKD progression, CV events, or both¹⁴.

To help policymakers and healthcare payers understand the financial implications of evidence-based interventions that reduce the risk of microvascular complications in patients with diabetes, we require a better understanding of the economic burden of both diabetes and microvascular complications, as well as up-to-date estimates of healthcare resource utilization (HRU) and medical costs.

The primary aim of this retrospective, observational, database study was to evaluate direct medical costs in patients with diabetes in the time periods before and after experiencing a microvascular complication, from the perspective of the Brazilian public healthcare system. A secondary aim was to explore associations between demographic or clinical characteristics and the incurrence of high direct medical costs in patients with diabetes who experienced a microvascular complication.

Methods

Detailed descriptions of the methodology and data sources utilized in this study have previously been published for an investigation of direct medical costs associated with macrovascular complications in Brazilian patients with diabetes¹⁷. Here we provide a summary of the methods used in the context of evaluating costs following a microvascular complication in Brazilian patients with diabetes.

Data source

The public health Brazilian Unified Health System (Sistema Único de Saúde [SUS]) is a free-to-patient service administered through a network of healthcare units, hospitals, laboratories, and blood banks, who are responsible for collecting, processing, and disseminating health information^18,19. Data collected by the SUS are stored in its database DATASUS²⁰, which comprises separate epidemiological, clinical, and administrative databases¹⁸. Additionally, all admissions to public hospitals are recorded in a National Hospital Information System (Sistema de Informações Hospitalares [SIH]), providing data on hospital admissions, discharges, and costs¹⁸. The Ambulatory Information System (Sistema de Informações Ambulatoriais [SIA]) stores data for public outpatient care covering emergency care, healthcare visits, and some high-complexity procedures (e.g. oncology). However, medical examinations and primary care data, including those related to medication use, are not publicly available in the database¹⁸. Anonymized data are freely available on the DATASUS website²⁰. Per the Brazilian ethical resolution number 510, ethical approval for this cost study was not required²¹.

Data collection

In the present study, we extracted data for eligible patients from the SUS-SIH and SUS-SIA databases between 1 January 2012 and 31 December 2018. The data extracted included administrative information (gender, age, region, hospital admission, and discharge date, healthcare unit, procedure, record date, and cost to the SUS) and medical information (diagnosis code[s]). Records for each individual were combined across these databases via record-linking methodology²².

Study population

Eligible patients had at least one record of diabetes (i.e. an International Classification of Diseases, 10th Revision [ICD-10] diagnosis code for diabetes, Table 1) and at least one microvascular complication (per ICD-10 diagnosis codes, Table 1) after the first record of diabetes during the study period. It should be noted that the translation of ICD-10 diagnosis codes in Brazil categorizes diabetes as insulin-dependent and non-insulin-dependent, distinct from and not interchangeable with type 1 and type 2 diabetes. Exclusion criteria included evidence of a record for cancer (per ICD-10 diagnosis codes for chemotherapy or radiotherapy); inconsistent data, which was defined as a negative age or unknown location; and an ICD-10 diagnosis code for a macrovascular complication during the study period (Table S1).

Table 1. ICD-10 diagnosis codes and frequency in patients with diabetes who experienced a microvascular complication (N = 2,096).

	Condition	ICD-10 code(s)	n (%)
Diabetes type	Insulin-dependent	E10, E10.0, E10.2, E10.3, E10.4, E10.5, E10.6, E10.7, E10.8, E10.9	632 (30.2)
	Non-insulin-dependent	E11, E11.0, E11.1, E11.2, E11.3, E11.4, E11.5, E11.6, E11.7, E11.8, E11.9	230 (11.0)
	Unspecified	E14, E14.0, E14.1, E14.2, E14.3, E14.4, E14.5, E14.6, E14.7, E14.8, E14.9	1,234 (58.9)
Microvascular complication	Diabetic foot including ulcer	L97	191 (8.4)
	Eye disease and diabetic retinopathy	H28.0, H36.0	117 (5.2)
	Kidney disease	N00.0, N00.3, N00.4, N00.8, N00.9, N03, N10, N11, N11.8, N11.9, N12, N15.8, N15.9, N16.3, N17, N17.0, N17.1, N17.8, N17.9, N18, N18.0, N18.8, N18.9, N19, Z49.1, Z94.0, Z99.2	1,667 (73.4)
	Neuropathy	G58.9, G59.0, G62.9, G63.2, M14.6	28 (1.2)
	Other microvascular	R02	267 (11.8)
	PVD	I78.9	0 (0.0)

Abbreviations. ICD-10, International Classification of Diseases, 10th Revision; PVD, peripheral vascular disease.

The denominator in the percentage calculations is 2,096 (the study cohort) for “diabetes type” and 2,270 (the qualifying number of microvascular complications) for “microvascular complications”. In total, 1,937 patients (92.4% of the study cohort), 145 (6.9%), 13 (0.6%), and 1 (0.0%) experienced 1, 2, 3, and 4 qualifying microvascular complications for study entry, respectively.

Study design

This analysis was conducted from the perspective of the Brazilian public healthcare system. The direct medical costs considered were to the SUS related to outpatient management, procedures, and hospitalization.

For each patient, the month of the qualifying microvascular complication was defined as the index date. Relative to this date, five study time periods were defined: baseline −1 (12–24 months before index), baseline (0–12 months before index), first-year (0–12 months after index), second-year (12–24 months after index) and third-year (24–36 months after index). All patients were required to have a minimum of two months’ data observable in the database following the index date.

The primary objective of this study was to describe the annual direct medical costs in patients with diabetes in the time periods before and after experiencing a microvascular complication. Annual direct medical costs were also described for patients stratified by gender and survival status in the overall study period. Information was available for a patient’s death if that event took place in the hospital or if the high-complexity procedures authorization (APAC) profile was closed due to death.

A secondary study objective was to explore potential associations between demographic or clinical characteristics and the incurrence of high direct medical costs, defined as the highest tertile of the annual costs ranked by median cost²³, in the overall study cohort. Additional analyses were conducted describing the length of hospital stay and the most frequent medical procedures and medications (based on patient numbers), defined per the SUS table of procedures, medications and orthoses, prostheses, and special materials (SIGTAP)²⁴. It should be noted that data for prescribed medications available in DATASUS are limited, and mainly include those prescribed in secondary and tertiary care.

Statistical analysis

All data were summarized descriptively, and missing data were not imputed. The costs were extracted from the source in Brazilian Real (BRL), then converted to 2019 USD using the average exchange rate of BRL 1: USD 0.252 for the third quarter of 2019 (1 July 2019–30 September 2019)²⁵. All costs reported here are in USD.

Potential associations between demographic or clinical characteristics and the incurrence of high direct medical costs were investigated using mixed-effects logistic regression analyses.

Analyses were performed in R (R Foundation for Statistical Computing, software version 3.6.1, Vienna, Austria).

Results

Study cohort

The patient flow is presented in Figure 1. There were 774,244 patients with evidence of diabetes in DATASUS during the 7-year study period¹⁷. Of these patients, 692,722 patients did not have evidence of a diabetes-related complication during the study period and were excluded from the study. After applying further exclusion criteria, there were 2,096 patients with diabetes who experienced 2,270 qualifying microvascular complications during the study period and were included in the overall study cohort. In total, 1,937 patients (92.4% of the study cohort), 145 (6.9%), 13 (0.6%), and 1 (0.0%) experienced 1, 2, 3, and 4 qualifying microvascular complications for study entry, respectively. The majority of the patients (n = 1,234; 58.9% of the study cohort) were categorized as having an unspecified diabetes type (Table 1). Insulin-dependent diabetes was the second most recorded (n = 632; 30.2%), followed by non-insulin-dependent diabetes (n = 230; 11.0%). Kidney disease accounted for most of the qualifying microvascular complications (n = 1,667; 73.4% of complications). Diabetic foot and eye disease were the next most common microvascular complications in the overall study cohort (n = 191; 8.4%, and n = 117; 5.2%, respectively). Events classified as “other” accounted for 267 (11.8%) of qualifying microvascular complications (Table 1).

Figure 1. Patient flow diagram. The overall study cohort included patients enrolled in DATASUS with evidence of diabetes and a microvascular complication during the study period who did not meet any exclusion criteria. ^†Defined as ≥1 record with an accepted ICD-10 diagnosis code for diabetes (Table 1); ^‡defined as ≥1 record with an accepted ICD-10 diagnosis code for a microvascular (Table 1) or macrovascular (Table S1) complication; ^§defined as ≥1 record with an accepted ICD-10 diagnosis code for a macrovascular complication (Table S1). Abbreviations. DATASUS, Ministry of Health’s Information Technology Department; ICD-10, International Classification of Diseases, 10th Revision.

Overall, the study cohort was 48.0% (1,006/2,096) female and the mean (standard deviation) duration of diabetes (time since the first evidence of diabetes during the study period and the qualifying microvascular complication) was 690 (595) days. More than half of the patients (n = 1,216; 58.0%) were from the southeastern region of Brazil (northeast: n = 366, 17.5%; south: n = 340, 16.3%; midwest: n = 129, 6.2%; north: n = 45, 2.1%). The median [interquartile range; IQR] age at the index date was 57.9 [49.5, 66.0] years. Median age ranged from 60.0 years for patients who experienced eye disease up to 69.0 years for those with diabetic foot (Table 2; groups not statistically compared). In total, 815 (38.9%) patients died during the overall study period.

Table 2. Age distribution of patients with diabetes who experienced a microvascular complication (N = 2,096).

Microvascular complication	n	Age at index date, years
		Mean ± SD	Median [IQR]	Minimum–maximum
Kidney disease	1,667	65.4 ± 14.8	66.0 [57.0; 76.0]	12.0–103.0
Other microvascular	269	67.8 ± 12.6	68.0 [59.0; 76.0]	27.0–99.0
Diabetic foot including ulcer	191	68.3 ± 12.1	69.0 [60.0; 76.5]	30.0–97.0
Eye disease and diabetic retinopathy	117	59.6 ± 13.1	60.0 [55.0; 69.0]	25.0–85.0
Neuropathy	28	62.7 ± 13.2	65.0 [55.0; 71.3]	30.0–81.0

ICD-10: International Classification of Diseases, 10th Revision; IQR: interquartile range; n: number of patients that contributed to data point; SD: standard deviation.

The month of the qualifying microvascular complication was defined as the index date.

Direct medical costs

Median [IQR] annual direct medical costs in USD/patient were 176.3 [91.0; 481.2] at baseline, increasing to 1,678.5 [287.0; 6,908.4] in the first year after a microvascular complication (Table 3; time periods not statistically compared). Median direct medical costs in USD/patient peaked in the second year (5,172.4 [274.8; 7,395.9]) and remained elevated above baseline in the third year (2,525.4 [167.0; 7,311.5]) after experiencing a microvascular complication. The general trends in direct medical costs were consistent in patients stratified by gender (Table 3; groups not statistically compared) or survival status during the overall study period (Table S2; groups not statistically compared). However, the increase in direct medical costs from baseline to the first year after a microvascular complication was more pronounced in patients who died during the study period than in those who survived, before reaching greater alignment in the second and third years (Table S2).

Table 3. Direct medical costs in patients with diabetes who experienced a microvascular complication, overall and stratified by gender.

Cohort	Time period	n	Direct medical cost, 2019 USD
			Mean ± SD	95% CI	Median [IQR]	Minimum–maximum
Overall (N = 2,096)	Baseline −1	975	310.2 ± 860.0	256.1; 364.2	113.7 [90.7; 236.8]	0.0–10,865.4
	Baseline	1,600	668.3 ± 1,438.7	597.7; 738.8	176.3 [91.0; 481.2]	0.0–15,940.0
	First year	2,096	3,425.1 ± 3,358.0	3,281.3; 3,569.0	1,678.5 [287.0; 6,908.4]	0.0–18,026.2
	Second year	1,178	4,261.6 ± 3,485.5	4,062.4; 4,460.9	5,172.4 [274.8; 7,395.9]	0.0–17,490.2
	Third year	943	3,674.0 ± 3,515.2	3,449.4; 3,898.7	2,525.4 [167.0; 7,311.5]	0.0–16,953.1
Female (n = 1,006)	Baseline −1	504	353.4 ± 1,007.2	265.2; 441.5	116.0 [90.7; 247.2]	0.0–10,865.4
	Baseline	776	757.2 ± 1,622.8	642.9; 871.6	181.4 [93.0; 535.6]	0.0–15,940.0
	First year	1,006	3,597.0 ± 3,383.7	3,387.6; 3,806.3	2,233.8 [306.6; 6,974.7]	0.0–18,026.2
	Second year	595	4,163.7 ± 3,441.4	3,886.6; 4,440.8	4,802.8 [277.6; 7,337.5]	0.0–15,454.9
	Third year	465	3,707.4 ± 3,548.4	3,384.1; 4,030.8	2,605.3 [154.0; 7,343.5]	0.0–11,286.5
Male (n = 1,090)	Baseline −1	471	263.9 ± 665.5	203.7; 324.2	110.9 [90.7; 222.6]	0.0–7,758.8
	Baseline	824	584.5 ± 1,235.8	500.0; 669.0	172.2 [91.0; 459.3]	0.0–15,596.3
	First year	1,090	3,266.5 ± 3,327.8	3,068.8; 3,464.3	1,324.7 [261.9; 6,868.8]	0.0–13,532.9
	Second year	583	4,361.6 ± 3,530.2	4,074.5; 4,648.8	5,553.8 [266.6; 7,446.9]	0.0–17,490.2
	Third year	478	3,641.5 ± 3,486.0	3,328.2; 3,954.9	2,444.9 [186.0; 7,298.0]	0.0–16,953.1

Abbreviations. CI, confidence interval; n: number of patients that contributed to data; IQR, interquartile range; SD, standard deviation; USD, United States dollars.

Baseline corresponds to the year before the qualifying complication, while baseline −1 refers to the year before baseline (i.e. between 1 and 2 years before the qualifying complication). First, second, and third year refer to the respective year after the qualifying complication. Costs were weighted by patient observation time using an annual general standardization.

When costs were stratified by cost category, costs were more evenly distributed between most categories at baseline than in the first year after a microvascular complication (Table S3; categories not statistically compared). At baseline, median costs (USD/patient) were in rank order as follows: clinical (166.3), transplants (156.4), surgical (151.2), orthoses (105.6), other (20.7), diagnostic (14.4), and medications (0.0). In the first year, the rank order of categories by median costs remained unchanged; however, median clinical costs increased by 2,129% to 3,707.6 USD/patient accounting for the vast majority of costs in the first year.

Figure 2 presents the results of exploratory mixed-effects logistic regression analyses assessing potential associations between demographic or clinical characteristics and the odds of incurring high direct medical costs. Time period, age, and region were all significantly associated with the incurrence of high direct medical costs. In comparison with baseline, odds ratios (ORs) [95% CI] for high direct medical costs were 69.92 [45.01; 108.61] in the first year and 84.68 [55.66; 128.82] in the second year (both p < .001). Likewise, as compared with patients aged over 80 years, those aged <40 years (OR 4.77 [95% CI: 2.21; 10.31]) or 40–<60 years (OR 2.93 [95% CI: 1.55; 5.52]) were more likely to incur high direct medical costs (both p < .001). There were also regional differences with greater odds of incurring high direct medical costs observed among residents of the midwestern (p = .007), north-eastern (p < .001), and southern (p < .001) regions compared with those from the south-eastern region.

Figure 2. Associations between demographic and clinical characteristics with the incurrence of high direct medical costs in patients with diabetes who experienced a microvascular complication (N = 2,096). Data are odds ratio [95% CI] for associations between parameters and high direct medical costs (defined as the highest tertile of annual cost data ranked by median cost for patients in the overall study cohort) from binary logistic regression modeling. Square brackets in the first column indicate the reference category. Baseline corresponds to the year before the qualifying complication, while first and second year refers to the respective year after the qualifying complication. Abbreviation. CI, confidence interval.

Length of hospital stay

Median hospital stays were 2 days at baseline, increasing to 3 days in the first year, before decreasing to 0 days in the second and third years after experiencing a microvascular complication (Table 4; time periods not statistically compared). Patient stays were of a broadly similar length by time period when data were stratified by gender (Table 4; groups not statistically compared).

Table 4. Length of hospital stay in patients who experienced a microvascular complication, overall and stratified by gender.

Cohort	Time period	n	Length of hospital stay per patient, days
			Mean ± SD	Median [IQR]	Minimum–maximum
Overall (N = 2,096)	Baseline −1	758	4.2 ± 14.0	0.0 [0.0; 4.0]	0.0–316.0
	Baseline	1,204	6.5 ± 16.7	2.0 [0.0; 8.0]	0.0–523.0
	First year	2,096	8.8 ± 15.7	3.0 [0.0; 11.0]	0.0–179.0
	Second year	472	3.4 ± 17.8	0.0 [0.0; 0.0]	0.0–546.0
	Third year	432	2.8 ± 12.0	0.0 [0.0; 0.0]	0.0–342.0
Female (n = 1,006)	Baseline −1	386	4.7 ± 16.4	0.0 [0.0; 4.0]	0.0–316.0
	Baseline	570	7.1 ± 21.4	2.0 [0.0; 8.0]	0.0–523.0
	First year	1,006	8.5 ± 15.7	3.0 [0.0; 10.0]	0.0–179.0
	Second year	229	3.1 ± 18.7	0.0 [0.0; 0.0]	0.0–546.0
	Third year	214	2.8 ± 13.3	0.0 [0.0; 0.0]	0.0–342.0
Male (n = 1,090)	Baseline −1	372	3.6 ± 11.1	0.0 [0.0; 3.0]	0.0–219.0
	Baseline	634	6.0 ± 10.8	2.0 [0.0; 8.0]	0.0–155.0
	First year	1,090	9.1 ± 15.6	3.0 [0.0; 12.0]	0.0–154.0
	Second year	243	3.6 ± 16.9	0.0 [0.0; 0.0]	0.0–320.0
	Third year	218	2.8 ± 10.8	0.0 [0.0; 0.0]	0.0–181.0

Abbreviations. n, number of patients that contributed to data point; IQR, interquartile range; SD, standard deviation.

Baseline corresponds to the year before the qualifying complication, while baseline −1 refers to the year before baseline (i.e. between 1 and 2 years before the qualifying complication). First, second, and third year refer to the respective year after the qualifying complication. Costs were weighted by patient observation time using an annual general standardization.

Medical procedures and medications

The three most frequent procedures (median total annual cost across all patients with the stated procedure in USD) for patients experiencing a microvascular complication were hemodialysis (n = 1,203; 2,043,407.2), treatment of diabetes (n = 1,595; 83,914.2), and confection of arterial fistula for hemodialysis (n = 699; 16,711.6) (Table 5). This was generally consistent across female and male patients (Table S4; groups not statistically compared). When the most frequent procedures were stratified by time period, hemodialysis procedures took place in 310 patients at baseline, rising to 1,111 patients in the first year after the complication; this was mirrored by the median hemodialysis cost per patient, which rose from 1,200.7 to 6,013.7 USD/year over the same time period (Table S5; time periods not statistically compared). When the most frequent procedures were stratified by both the time period and the qualifying microvascular complication, the procedures tended to align with those that are associated with the complication (Table S6). For instance, those associated with kidney complications were frequently observed in patients who experienced kidney disease, and those associated with eye examinations were common in patients with eye disease and diabetic retinopathy.

Table 5. Direct medical costs for the ten most frequent medical procedures and medications for patients with diabetes who experienced a microvascular complication (N = 2,096).

Description	n	Total annual direct medical cost, 2019 USD
Medical procedures
Associated with kidney complications
Hemodialysis (≤3 sessions per week)	1,203	2,043,407.2
Confection of arteriovenous fistula for hemodialysis	699	16,711.6
Long-term catheter implant for hemodialysis	450	16,622.3
Post-transplant follow-up (kidney, liver, heart, lung, hematopoietic stem cells, and/or pancreas transplant)	34	3,672.5
Associated with peripheral vascular disease
Finger amputation/disarticulation	370	8,541.2
Amputation/disarticulation of lower members	318	20,308.1
Associated with diabetes, other microvascular complications and other
Treatment of diabetes mellitus	1,595	83,914.2
Treatment of complicated diabetic foot	579	15,878.6
Maintenance and household accompanying of patients	175	207,546.2
Home hospitalization	26	5,322.2
Medications
Associated with kidney complications
Ferric hydroxide saccharate 100 mg injectable (per 5 mL bottle)	649	3,759.0
Epoetin alfa 4,000 IU injectable (per bottle)	628	901.7
Calcitriol 0.25 mcg (per capsule)	248	770.9
Sevelamer 800 mg (per pill)	273	347.2
Calcitriol 1.0 mcg injectable (per ampoule)	99	443.6
Epoetin alfa 3,000 IU injectable (per bottle)	89	1,474.9
Epoetin alfa 2,000 IU injectable (per bottle)	62	48.4
Sodium mycophenolate 360 mg (per pill)	27	0.0
Associated with other microvascular complications
Atorvastatin 20 mg (per pill)	36	0.0
Atorvastatin 10 mg (per pill)	23	0.0

Abbreviations. IU, international units; n, number of patients that contributed to data point; SUS, Sistema Único de Saúde (the Brazilian public health system).

Data presented are total annual direct medical costs (2019 United States dollars) across patients (with the stated procedure or medication) over the study period, both before and after the qualifying complication. Costs were weighted by patient observation time using an annual general standardization. Procedures and medications were defined as per the SUS table of procedures, medications and orthoses, prostheses, and special materials [24].

The medications prescribed in mostly secondary and tertiary care settings accounted for a relatively small share of annual direct medical costs across the study period (Table S3). The most commonly utilized medications were ferric hydroxide saccharate and high-dose epoetin alfa (4,000 IU), both used in the management of anemia in patients with CKD, and which added modest direct medical costs (Table 5). The patterns of medication use were broadly similar among men and women (Table S7; groups not statistically compared). Medications were rarely used at baseline, with the most frequent (atorvastatin 10 mg) prescribed to eight patients (Table S8). Medication use dramatically increased in the first year after experiencing a microvascular complication, remaining elevated above baseline levels in the second and third years, while accounting for modest per-patient costs.

Discussion

This retrospective database study, from the perspective of the Brazilian public healthcare system, has demonstrated a large increase in direct medical costs in patients with diabetes after experiencing a microvascular complication. The costs were highest in the second year after experiencing a microvascular complication and remained elevated above baseline over three years (after which the study period ended).

Exploratory analyses suggested that the incurrence of high direct medical costs was strongly associated with the time period, with much higher odds in the first (OR 70) and second (OR 85) year after experiencing a microvascular complication, in comparison with baseline. There were some smaller regional differences in the odds of incurring high medical costs, as well as significant differences between some of the younger age groups when compared with the oldest patients. The geographical factors associated with an increased likelihood of high direct medical costs (Figure 2) may be partly explained by regional differences in access to tertiary care and diabetologists, as well as by the movement of patients between regions to access treatment.

These findings are in line with and add to our previous demonstration of increased direct medical costs associated with macrovascular complications in Brazilian patients with diabetes¹⁷. Moreover, it should be noted that the odds of patients incurring high direct medical costs in the first and second year after the complication (vs. baseline) were substantially greater for microvascular complications (ORs of 70–85; USD/patient of 1,679 [first year] and 5,172 [second year]) here than for macrovascular complications with or without a microvascular complication (ORs of 3–6; USD/patient of 334 [first year] and 266 [second year]) in our previous study¹⁷. One potential explanation for this difference is the substantial rise in costs as nephropathy progresses²⁶; the management of patients with end-stage kidney disease consumes considerable healthcare resources, requiring long-term dialysis or kidney transplantation²⁷. Kidney disease was by far the most common qualifying microvascular complication in our study and a sizable proportion of this cohort received hemodialysis in the first and second years after experiencing the complication. This is perhaps not surprising, as DATASUS does not cover primary care, only secondary and tertiary care.

The observation of higher costs associated with microvascular vs. macrovascular complications is in agreement with findings from a cohort of patients with type 2 diabetes in China²⁸, but the contrast with those from studies of type 2 diabetes in Brazil²⁹ and Europe³⁰, where macrovascular complications were associated with the higher costs. The reasons for these contrasting findings are not entirely clear, but likely reflect important differences in the analysis and estimation of costs and between the patient populations studied.

Our results add to the growing evidence of the substantial economic burden on healthcare systems worldwide due to diabetes-related complications^{8,9,12,17,31,32}. The high cost of managing diabetes represents a growing challenge for healthcare systems. For example, Bahia et al. estimated the total economic burden of diabetes in Brazil to be 15.7 billion international dollars from a societal perspective in 2014, i.e. 0.5% of Brazilian gross domestic product³¹. In 2019, the International Diabetes Federation estimated that nearly 1 in every 4 BRL (24.2%) of total healthcare expenditure was attributed to diabetes¹.

The ongoing COVID-19 pandemic may further increase the economic burden due to diabetes and its complications. There is a widely recognized association between diabetes and the severity of COVID-19 symptoms, including the risks of admission to intensive care and mortality^33–35. Therefore, the current pandemic is likely to increase diabetes-related direct medical costs, exacerbating the burden on healthcare systems. The results presented here add further evidence, specifically in a Brazilian public healthcare setting, that microvascular complications substantially contribute to direct medical costs in patients with diabetes. Accordingly, our findings provide support for the need to appraise and implement evidence-based and cost-effective strategies that help to reduce the burden of diabetes-related microvascular complications. Such strategies may include those for optimizing lifestyle management³⁶, addressing the non-uniformity of access to and availability of essential medicines within the SUS^37,38, for example through programs, such as the Pharmacy Network of Minas (Rede Farmácia de Minas [RFM])³⁹, and strategies to improve sub-optimal treatment adherence among patients with diabetes⁴⁰.

Our study has some limitations that should be highlighted. Originally, the SUS-SIH and SUS-SIA were created for administrative purposes, meaning the data source has limited clinical information and may be prone to errors. Furthermore, the DATASUS database does not cover primary care, only secondary and tertiary care. Therefore, there is a risk of selection bias in the data sample, as most patients with diabetes in Brazil will be seen in primary care and those patients with complications are likely to be transferred to secondary and/or tertiary care. It should also be noted that DATASUS is a reimbursement database, which may have an impact on the types of microvascular complications reported herein. Additionally, as with other retrospective database studies, our results are reliant on the completeness of the records, which in our case may not contain all relevant or complete information. For instance, the majority of patients are classified as having “unspecified diabetes”, rather than insulin-dependent or non-insulin-dependent disease (Table 1). Accordingly, we have minimized attention on diabetes type and concentrated on analyses in patients with diabetes (any type) according to complication status, gender, and time period. Our analyses did not include assessment of costs stratified by additional incidences of microvascular complication and this is an area for further study. While we excluded from this analysis patients with a microvascular complication who also had a macrovascular complication, medical costs data for this subgroup is reported in our recent article focusing on macrovascular complications in Brazilian patients with diabetes¹⁷.

It is worth noting that the translation of ICD-10 diagnosis codes for diabetes varies between Brazil and other countries (e.g. the United Kingdom and the United States), which categorize type 1 and type 2 diabetes; these distinctions (insulin dependence vs. diabetes type) are not interchangeable and we recognize this as a study limitation. A related limitation is the different use in Brazil relative to other countries of ICD-10 codes to identify microvascular complications. Also, the diagnosis coding system may not exclusively distinguish between a microvascular and macrovascular, or mixed, origin for complications, such as foot ulcers, and this is a further limitation of our work. We also acknowledge that direct medical costs in stratified patient groups may have been influenced by differences between groups in mortality rates. A final limitation is that our findings may not be representative of the entire Brazilian population given that ∼75% of the population of Brazil depend exclusively on the SUS, while the remaining 25% have some access to private healthcare, through which treatment for diabetes and its complications may, at least in part, be provided. In using data from the SUS for our analyses, we thus could not quantify the direct medical costs within the private healthcare sector, where the standards of care and costs are likely to be higher compared with the SUS¹⁹.

Conclusions

The present study has demonstrated higher medical costs in patients with diabetes after experiencing a microvascular complication when compared with the costs incurred before the complication. In addition to providing up-to-date cost estimates, our results highlight the need to appraise the cost-effectiveness of evidence-based strategies to reduce the risk of microvascular complications in patients with diabetes in Brazil.

Transparency

Declaration of funding

Sponsorship for this analysis and article processing charges were funded by Novo Nordisk A/S.

Declaration of financial/other interests

At the time of the study, GSJ and DC were full-time employees of IQVIA Brasil, which received consultancy fees from Novo Nordisk A/S to conduct the analyses. GSJ remains a full-time employee of IQVIA Brasil, while DC has left the company. LMAP and RCLAC are employees of Novo Nordisk Farmacêutica Do Brasil Ltda. NM and JBH are employees of Novo Nordisk A/S. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

GSJ is the guarantor of this work and, as such, had full access to all the data in the study, and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors confirm that they meet the International Committee of Medical Journal Editors (ICJME) uniform requirements for authorship and that they have contributed to the CRediT roles detailed below. GSJ: data curation; formal analysis; methodology; validation; visualization; writing – review and editing. RCLAC: writing – review and editing. NM: conceptualization; methodology; supervision; writing – review and editing. JBH: methodology; writing – review and editing. LMAP: writing – review and editing. DC: data curation; formal analysis; methodology; validation; visualization; writing – review and editing. All authors and share in the final responsibility for the content of the manuscript, as well as the decision to submit it for publication.

Geolocation information

Brazil.

Acknowledgements

The authors would like to thank the participants of the study and acknowledge the input of Flávia Sauer Tobaruella to early discussions. The medical writing support was provided by David Floyd on behalf of, and editorial assistance was provided by Beverly La Ferla of, Ashfield MedComms, an Ashfield Health company, funded by Novo Nordisk A/S.

Data availability statement

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.