Abstract
Objective
Diabetic foot ulcer, which often leads to lower limb amputation, is a devastating complication of diabetes that is a major burden on patients and the healthcare system. The main objective of this study is to determine the economic burden of diabetic foot ulcer-related care.
Methods
We conducted a multicenter study of all diabetic foot ulcer patients admitted to general internal medicine wards at seven hospitals in the Greater Toronto Area, Canada from 2010 to 2015, using the GEMINI database. We compared the mean costs of care per patient for diabetic foot ulcer-related admissions, admissions for other diabetes-related complications, and admissions for the top five most costly general internal medicine conditions, using the Ontario Case Costing Initiative. Regression models were used to determine adjusted estimates of cost per patient. Propensity-score matched analyses were performed as sensitivity analyses.
Results
Our study cohort comprised of 557 diabetic foot ulcer patients; 2939 non-diabetic foot ulcer diabetes patients; and 23,656 patients with the top 5 most costly general internal medicine conditions. Diabetic foot ulcer admissions incurred the highest mean cost per patient ($22,754) when compared to admissions with non-diabetic foot ulcer diabetes ($8,350) and the top five most costly conditions ($10,169). Using adjusted linear regression, diabetic foot ulcer admissions demonstrated a 49.6% greater mean cost of care than non-diabetic foot ulcer-related diabetes admissions (95% CI 1.14–1.58), and a 25.6% greater mean cost than the top five most costly conditions (95% CI 1.17–1.34). Propensity-scored matched analyses confirmed these results.
Conclusion
Diabetic foot ulcer patients incur significantly higher costs of care when compared to admissions with non-diabetic foot ulcer-related diabetes patients, and the top five most costly general internal medicine conditions.
Introduction
Diabetic foot ulcers (DFUs) are one of the most common complications of diabetes, with incidence ranging between 2.2% and 6.0% across different populations.1–4 It is estimated that up to 26.1 million adults are expected to develop foot ulcers per annum worldwide. 4 Moreover, foot ulceration confers a high-risk prognosis for lower limb loss and mortality. Nearly 20% of DFUs will eventually result in amputation, and DFUs carry a two-fold higher risk of mortality when compared to patients with other diabetes-related complications.4–6
Due to the high incidence rate and amputation risk associated with DFUs, treatment is expected to be costly. Numerous studies have assessed the costs associated with DFU-related care across varying populations and healthcare settings. Despite differences in study designs, patient population, study setting, foot ulcer severity, treatment methods, and differing accountings of direct and indirect costs, among other differences, treatment costs for DFU-related care are consistently reported to be high.7–15
Although there is an abundance of data regarding the cost associated with DFU care, these costs have not been sufficiently characterized in the context of the single-payer, publicly funded, Canadian healthcare system. Moreover, a single-payer healthcare setting ensures that all patients have access to adequate and proper care for their conditions. Hence, this gives us a unique opportunity to compare the cost associated with DFU-related care to other conditions. Consequently, we set out to determine the economic burden of inpatient DFU treatment in Toronto, Canada.
Methods
Study design, setting and data sources
We conducted a multicenter, retrospective cohort study of all general internal medicine (GIM) patients admitted to Toronto hospitals with admitting diagnoses of DFU from January 2010 to March 2015. Toronto is Canada’s most populous and multicultural city with over 2.5 million residents. The General Medicine Inpatient Initiative (GEMINI) database, which electronically collects information on all GIM admissions at five teaching and two community hospitals in Toronto, Ontario, Canada, served as the data source. The patients captured in GEMINI come from all socioeconomic strata, with their hospital care fully funded within the single-payer public health system of the province of Ontario. The seven GEMINI hospitals experienced 136,208 admissions to their GIM ward, and their GIM admissions accounted for approximately 39% of all emergency department admissions. This database has previously been described in detail. 16
Patient cohort
The patient cohort included DFU patients and two comparator groups from the GEMINI database. Patients with a most responsible diagnosis of DFU served as the intervention group and were identified using the Canadian edition of the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10-CA) codes (Table S1 in the Supplementary section). The first comparator group consisted of GIM inpatients with other diabetes-related complications, such as coma, ketoacidosis, lactic acidosis, ketoacidosis and lactic acidosis together, incipient diabetic neuropathy, advanced kidney diseases, background retinopathy, preproliferative retinopathy, proliferative retinopathy, cataract, advanced ophthalmic disease, iritis, retinitis, mononeuropathy, polyneuropathy, peripheral angiopathy, peripheral angiopathy with gangrene, diabetic arthropathy, diabetic hand syndrome, and hypoglycemia, among others (refer to ICD-10-CA codes in Table S1 of the Supplementary section). The second comparator group included inpatients with any of the top five most costly GIM conditions. Patients with the top five most costly conditions were identified and defined using a similar methodology to that previously described. 17 However, we enrolled patients according to the most costly conditions rather than the most common conditions. The five most costly conditions were identified as congestive heart failure, pneumonia, urinary tract infection, chronic obstructive pulmonary disease, and pneumonitis. For a full list of codes used to define the diabetes and non-diabetes groups, refer to Table S1 in the Supplementary section.
Outcomes
The primary outcome of this study was the mean cost per inpatient admission. Costs were estimated using the Ontario Case Costing Initiative (OCCI). This includes the costs at the patient/resident level for day surgeries, ambulatory care cases, acute inpatient admissions, mental health visits, and complex continuing care, among others. 18 Cost data were available for all GEMINI hospitals except one. Secondary outcomes for this study included in-hospital amputation rate (major vs. minor vs. no amputation), in-hospital length of stay (days), patient disposition (to either acute care hospital, against medical advice, continuing care, death, home, or otherwise), overall and 30-day readmission rates to GIM at one of the GEMINI hospitals (%), in-hospital major adverse cardiovascular events (MACE – defined as stroke, myocardial infarction (MI], or death), and all-cause mortality (%), within the study’s duration (2010–2015). Event outcomes such as MACE and amputation rate were captured using ICD-10-CA codes (refer to Table S1 in the Supplementary section).
Covariates
Baseline characteristics including admission year, age, sex, medical comorbidities, previous peripheral revascularization, and dialysis were collected for each patient (for full list, refer to Table 1). Hospital specific variables included volume of GIM admissions and hospital teaching status. Stroke was defined as previously described. 17 The remaining comorbidities, as well as peripheral revascularization status were defined using ICD-10 codes. Dialysis was identified using Canadian Classification of Health Interventions (CCI) codes (see Table S1). The presence of any comorbidity was defined as the presence of any of the listed codes as secondary diagnoses on the index admission.
Baseline characteristics of the overall patient cohort, before and after matching.
Significant values are noted with an asterisk.
Difference among group means was tested using analysis of variance (one-way ANOVA).
bComparison between study groups done using Chi-square test.
DFU: diabetic foot ulcer; MI: myocardial infarction.
Statistical analysis
Demographic data were expressed as means with standard deviations (SD) or counts and percentages. Evaluation of baseline characteristics between DFU patients and comparator groups was performed using one-way analysis of variance (ANOVA) for continuous variables, and Fisher’s exact or Chi-square tests for categorical variables. A generalized linear model with a gamma distribution and a log-link function was used to analyze difference in healthcare costs between groups. 19 Negative binomial regression was utilized for the length of hospital stay (LOS) in order to account for over dispersion. Amputations, patient disposition, 30-day re-admission, MACE and death were modeled with logistic regression. Estimates of total healthcare costs comparing DFU admissions with non-DFU and top five most costly conditions admissions were reported as the exponentiated coefficient. Incidence rate ratios were reported for LOS, and odds ratios were reported for amputations, patient disposition, 30-day re-admission, MACE and death. Subgroup analyses were conducted measuring the total cost of care stratified by amputation type, MACE, and mortality across all three patient populations.
Propensity score matching was performed using the three study groups (DFU, non-DFU and top 5 most costly conditions) in order to control for confounders, improve the homogeneity of the case mix, and ultimately identify comparable groups for analysis. A logistic regression model was utilized to estimate the propensity score for each patient, defined as the probability of DFU conditional on selected covariates. Confounding variables included age, gender, dialysis, acute myocardial infarction, coronary artery disease, chronic kidney disease, cerebrovascular disease, peripheral artery disease. Propensity score matching was done using nearest neighbor matching without replacement, in which a matched cohort was constructed where each DFU patient was matched to the closest non-DFU and top five most costly conditions patients with a ratio of 1:1:2. A greedy algorithm was implemented to match those whose estimated propensity score logits were within 0.2 propensity score logit standard deviations from each other.20–22 Balance assessment, for variables used in the propensity score matching, was performed using absolute standardized differences. Imbalance was defined as a standardized difference greater than 0.1 in absolute value. 20 , 21 The matched groups were compared on the study outcomes using a multivariable regression analysis, adjusting for variables that were still imbalanced after the matching. In sensitivity analysis, the unmatched cohort, including all available patients, was used in a multivariable backward selection model to adjust for confounding.
All study analyses were conducted in R studio, version 3.5.1 (R Foundation for Statistical Computing), and were carried out at a 5% two-sided significance level.
Results
Our cohort comprised of 27,152 (51.1% male) admissions in total. Diabetic foot ulcers accounted for 557 (2.1%) patients, whereas patients with other diabetes-related complications and the top five most costly conditions accounted for 2939 (10.8%) and 23,656 (87.1%) patients, respectively. With respect to baseline characteristics, male sex, acute MI, chronic kidney disease, dialysis, cerebrovascular disease, and peripheral arterial disease were more prevalent among the DFU group, whereas age and coronary artery disease were found to be significantly higher in the top five most costly conditions group. Table 1 illustrates baseline characteristics for the unmatched and matched cohorts, whereas absolute standardized differences comparing baseline covariates between the unmatched and matched cohort are reported in the Supplementary Figure 1.
Unadjusted outcomes analysis
In the unadjusted cohort, DFU admissions incurred the highest mean cost per inpatient admission ($22,754, SD 32,066) in comparison to other diabetes-related admissions ($8351, SD 18,154) and the top five most costly conditions ($10,169, SD 23,912) (p≤0.001). DFU patients also had a significantly higher amputation rate than non-DFU diabetes patients and the top five most costly admissions (21.9% vs. 0.3% vs. 0.0%, respectively, p ≤0.001). Foot ulcerations also resulted in the longest mean length of in-hospital stay when compared to patients with other diabetes-related complications and the top five most costly conditions (17 vs. 7 vs. 9 days, p≤0.001).
With respect to patient disposition, DFU patients had the highest proportion of disposition to acute care hospital (1.3%) and rehabilitation homes (28%), whereas the non-DFU diabetes group had the highest percentage of patients that were sent home (81.6%) or discharged against medical advice (1.4%). Rates of readmission and 30-day readmission were found to be similar across all three groups. Mortality and MACE rates were highest for the top five most costly conditions group (Table 2).
Unadjusted primary and secondary outcomes, compared between all three groups, before and after matching.
*Significant values are noted with an asterisk.
aDifference among group means was tested using analysis of variance (One-way ANOVA).
bComparison between study groups done using Chi-square test.
DFU; diabetic foot ulcer.
Regression analyses
After adjusting for baseline differences, DFU-related care was found to be 38.9% (exp(β) 1.389, 95% CI 1.336–1.445) more costly when compared to non-DFU diabetes patients, and 21.5% (exp(β) 1.215, 95% CI 1.166–1.265) more costly than patients in the top five most costly conditions group. For patients that did not undergo an amputation, DFU admissions incurred 25.7% (exp(β) 1.257, 95% CI 1.218–1.296) and 9.3% (exp(β) 1.093, 95% CI 1.061–1.126) greater mean costs of care when compared to patients in the non-DFU diabetes and the top five mostly conditions groups, respectively. In-hospital length of stay was also found to be 33.4% (exp(β) 1.334, 95% CI 1.283–1.474) longer for DFU patients than patients with other diabetes-related complications, and 22.5% (exp(β) 1.225, 95% CI 1.177–1.276) longer than for patients in the top five most costly conditions group. Discharge to continuing care was significantly more likely for DFU patients (Table 3).
Unadjusted and adjusted regression analysis of the overall cohort, with the top five most costly conditions as the reference group.
Significant values are noted with an asterisk.
aExponentiated regression coefficient as the continuous outcome was log transformed.
bOdds ratio for logistic regression models used.
cNon-DFU was the reference group.
Multivariate analyses of the primary and secondary outcomes.
Significant values are noted with an asterisk.
aAdjusted for age, gender, dialysis, pneumonia.
CHF: congestive heart failure; CKD: chronic kidney disease; CVD: cerebrovascular disease; PAD: peripheral artery disease; UTI: urinary tract infection.
Propensity score matched analysis
Analysis of the unadjusted, matched cohort confirmed the significantly higher treatment costs associated with DFUs when compared to patients with other diabetes-related complications and the top five most costly conditions ($22,890 vs. $8029 vs. $9249, p≤0.001). The remaining outcomes were almost identical to those in the overall cohort (Table 2).
After adjusting for confounding factors, our propensity matched sensitivity analysis demonstrated a 49.6% higher mean cost of care for the diabetic foot ulcer patients compared to the non-DFU diabetes group (exp(β) 1.496, 95% CI 1.14–1.58) and a 25.6% greater mean cost for the diabetic foot ulcer patients compared to top five most costly condition group (exp(β) 1.256, 95% CI 1.17–1.34). The log transformed cost of care for patients that did not undergo amputations in the DFU group was greater than for patients with other diabetes-related complications (exp(β) 1.39, 95% CI 1.34–1.45), as well as the top five most costly conditions (exp(β) 1.12, 95% CI 1.06–1.17) (Table 4). Moreover, diabetic foot ulcer patients had a significantly higher length of hospital stay (days) when compared to patients with other diabetes-related complications, and the top five most costly conditions. Discharge to continuing care was also highest among DFU patients.
Subgroup analyses
Cost of care for patients that underwent a minor or major amputation was observed to be marginally higher for patients with other diabetes-related complications than with patients in the DFU group, although this difference was not statistically significant. However, for patients that did not undergo any amputations, treatment costs were significantly higher for the diabetic foot ulcer group ($17,182, SD 18,994), when compared to the non-DFU diabetes group ($8231, SD 17,996) and top five most costly conditions ($10,169, SD 23,912) (p ≤ 0.0001).
We also observed that MACE significantly augmented the treatment costs more for foot ulcer patients ($65,167) than patients with other diabetes-related complications ($24,694) and the top five most costly conditions ($16,334) (Supplementary Figure 2). Furthermore, DFU patients with a secondary diagnosis of MACE demonstrated a 29.3% (95% CI 1.114–1.5098) greater total cost of care than patients admitted with the top five most costly conditions (p ≤ 0.0001). Similarly, among the subgroup of patients with mortality, this phenomenon was also observed when the cost pertaining to mortality was assessed, although these differences were not significant (Supplementary Figure 3).
Discussion
In this large multicenter study, we observed that treatment costs were significantly higher for diabetic foot ulcer inpatients when compared to patients with other diabetes-related complications and the top five most costly causes of GIM admission. Amputation rate, mean in-hospital length of stay, and discharge to continuing care were highest for DFU patients. Among patients stratified by amputation status, DFU patients without amputations incurred the most costs. Moreover, the cost associated with MACE and mortality was also found to be higher for diabetic foot ulcer patients. Compared to the data published by Verma et al., 17 DFUs ($22,754) were found to have a higher median cost than each individual disease in the top five most costly conditions group: congestive heart failure ($6,717), pneumonia ($6,009), chronic obstructive pulmonary disease ($6,148), urinary tract infection ($5,071), and pneumonitis ($8,350). 17
Numerous studies from around the world have investigated the costs associated with treating diabetic foot ulcers across different healthcare systems. Tennvall et al. 15 estimated the median cost for managing deep foot infections in Swedish patients with diabetes to be SEK 206,200 (approximately $36,000 USD using 1997 price levels used in the study). However, enrollment for this study was limited to one hospital. In contrast, our study recruited patients from seven different treating centers. Another study, by Ramsey et al., 23 reported the two-year attributable cost (post-diagnosis) for diabetic foot ulcer care to be $28,000 USD. However, a major limitation of this study was that the authors were not able to determine if amputations were a result of diabetic foot ulcers, or some other conditions. The current study was able to overcome this limitation by only including patients that had a most responsible diagnosis of foot ulcerations. Thus, we can be more certain that it was the diabetic foot ulcer diagnosis that led to amputations in our study. Furthermore, Chan et al. 10 conducted a recent systematic review assessing the cost pertaining to different chronic ulcers, in which they reported mean cost per hospital episode of $16,400 USD for treating diabetic foot ulcers (from the health-care public payer perspective). However, only two studies met their inclusion criteria for this specific outcome. Other studies have reported slightly lower treatment costs associated DFU-related care, ranging from $3600 to $28,000. 9 ,24–26 Despite the methodological differences, the cost pertaining to DFU-related care is consistently reported to be high across different healthcare systems.
With regard to Canada specifically, the cost associated with the treatment of diabetic foot ulcers has been insufficiently studied. In one of few available studies, O’Brien et al. reported an average, unadjusted treatment cost of $2,183 per diabetic foot ulcer episode – a stark contrast to our mean, unadjusted cost of $22,754 per DFU case. This difference may be attributed to O’Brien et al. 27 defining foot ulcers as “those that heal without amputation or vascular surgery,” thus omitting foot ulcers leading to amputation. In contrast, we were able to calculate treatment cost for DFU patients that did and did not undergo amputations. Moreover, O’Brien et al. 27 used only one ICD-9 code to identify their DFU patient cohort, whereas we used a few ICD-10 codes to identify our DFU cohort. In a separate study, Hopkins et al. 28 estimated the annual cost for DFU-related care, per prevalent Canadian case, to be $21,371. This was very similar to our mean, unadjusted, cost of $22,754 for DFU-related care. This slight difference in estimated treatment cost may be attributed to the fact that Hopkins et al. utilized ICD-10 codes that combined diabetes with decubitus ulcer as a method to identify some of their DFU patients. This is a domain of disagreement between us, as we do not deem decubitus ulcer to be an accurate diagnosis of diabetic foot ulcers. In addition, Hopkins and authors failed to account for attribution bias, as no matching or multivariate analyses were performed. In the current study, we performed regression analysis on the overall unmatched and propensity-score matched cohorts, ensuring our results were robust. This further strengthens the notion that DFU-related care in Canadian patients is costly. To the author’s best knowledge, there are no more published studies that have investigated the cost of DFU-related care in Canada.
From prior knowledge of the literature and meticulous statistical analyses, we hypothesize that a few factors drive up the costs for DFU-related care. First, amputations are known to drive up cost of care for any diagnosis. For instance, the cost for limb ischemia patients undergoing amputations is reported to be $24,700 by Raviola et al. and $26,142 by Mackey et al. 29 , 30 For diabetic foot lesions specifically, some studies estimate the cost of amputation to range from approximately $15,000 to up to $67,000.31–36 In the current study, amputation cost was higher in patients with other diabetes-related complications than the diabetic foot ulcer group, although this did not reach statistical significance. The most likely explanation of this phenomenon is that they were only eight patients (0.3%) in the non-DFU diabetes group that underwent a minor or major amputation, in comparison to 122 (22%) from the DFU group. It is likely that this small sample size of eight patients was more prone to random error. Furthermore, amputation patients are often required to stay in the hospital for longer periods of time, as they often need time to recover from surgery, treat wound complications, and undergo a waiting period for rehabilitation in order to re-learn how to walk, among other reasons. This utilizes further healthcare resources, which in turn augments the treatment cost. Because DFU patients (21.9%) had a significantly higher amputation rate than non-DFU diabetes patients (0.3%) and the top five most costly conditions (0.0%), their utilization of other healthcare resources is much higher.
However, other factors also drive up treatment costs for DFU-related care, as we found treatment costs for patients that did not undergo amputations to be significantly higher for the DFU group, when compared to the non-DFU diabetes group and the top five most costly conditions as well ($17,182 vs. $8,231 vs. $10,169, respectively). A plausible explanation of this is the increased need and usage of wound care for patients with diabetic foot ulcers, as studies have demonstrated the high costs incurred by topical treatments with regard to DFU-related care. 7 , 31 , 34
There are several limitations to this study. First, we were not able to capture patients that were admitted to non-general internal medicine services. In particular, patients with peripheral arterial disease (PAD) presenting to vascular surgery clinics (who have a high rate of DFUs) were not captured. Hence, our sample size is an underestimation of the total number of patients that are admitted with a diabetic foot ulcer. Secondly, the ICD-10 codes used in this study to identify the patient population have not been validated, to the best of our knowledge. Thirdly, we were unable to capture the DFU classification system used by the diagnosing physician, as well as other-related costs such as wound debridement. Lastly, database limitations prevented us from capturing out-of-hospital costs pertaining to diabetic foot ulcers, such as the purchase of special equipment (power wheelchairs, accessible ramps for patient homes, and modified steering wheels, among others). If accounted for, these out-of-hospital costs would likely increase the economic burden of diabetic foot ulcers on the Canadian healthcare system, as some of these are partially subsidized by the government.
With that being said, our study also has strengths that warrant discussion. This study is the first Canadian comparative study to date to investigate the economic burden associated with the care of DFUs and non-DFU diabetes and the top five most costly conditions. Second, we were able to investigate previously unstudied outcomes for patients with a DFU, such as patient disposition and rate of major adverse cardiovascular events. Third, we performed a variety of statistical analyses to ensure robustness and accuracy of our results. We adjusted for confounding variables in multivariate regression models, and also conducted propensity-matched analysis. Fourth, because our data were extracted and collected from seven different hospitals in Toronto and consisted of 27,152 patients, our findings and results can be considered geographically generalizable.
In summary, we found that DFU patients have a significantly higher cost of care and utilization of healthcare resources when compared to patients with other diabetes-related complications, and patients with the top five most costly conditions. This highlights the need and necessity for focused efforts on foot ulcer prevention strategies and informs health policy in that regard.
Supplemental Material
sj-pdf-1-vas-10.1177_1708538120923420 - Supplemental material for The economic burden of inpatient diabetic foot ulcers in Toronto, Canada
Supplemental material, sj-pdf-1-vas-10.1177_1708538120923420 for The economic burden of inpatient diabetic foot ulcers in Toronto, Canada by Muzammil H Syed, Konrad Salata, Mohamad A Hussain, Abdelrahman Zamzam, Charles de Mestral, Mark Wheatcroft, John Harlock, Deana Awartani, Badr Aljabri, Amol Verma, Fahad Razak, Subodh Verma and Mohammed Al-Omran in Vascular
Footnotes
Acknowledgements
The article has been presented at the World Congress of Cardiology and Cardiovascular Health 2018, hosted by the World Heart Federation on 5–8 December 2018, in Dubai, United Arab Emirates.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was partially funded by the Society of Vascular Surgery Student Research Fellowship
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
