OBJECTIVE

To estimate the rates of diabetes complications and revascularization procedures among people with diabetes who have experienced homelessness compared with a matched cohort of nonhomeless control subjects.

RESEARCH DESIGN AND METHODS

A propensity-matched cohort study was conducted using administrative health data from Ontario, Canada. Inclusion criteria included a diagnosis of diabetes and at least one hospital encounter between April 2006 and March 2019. Homeless status was identified using a validated administrative data algorithm. Eligible people with a history of homelessness were matched to nonhomeless control subjects with similar sociodemographic and clinical characteristics. Rate ratios (RRs) for macrovascular complications, revascularization procedures, acute glycemic emergencies, skin/soft tissue infections, and amputation were calculated using generalized linear models with negative binomial distribution and robust SEs.

RESULTS

Of 1,076,437 people who were eligible for inclusion in the study, 6,944 were identified as homeless. A suitable nonhomeless match was found for 5,219 individuals. The rate of macrovascular complications was higher for people with a history of homelessness compared with nonhomeless control subjects (RR 1.85, 95% CI 1.64–2.07), as were rates of hospitalization for glycemia (RR 5.64, 95% CI 4.07–7.81) and skin/soft tissue infections (RR 3.78, 95% CI 3.31–4.32). By contrast, the rates of coronary revascularization procedures were lower for people with a history of homelessness (RR 0.76, 95% CI 0.62–0.94).

CONCLUSIONS

These findings contribute to our understanding of the impact of homelessness on long-term diabetes outcomes. The higher rates of complications among people with a history of homelessness present an opportunity for tailored interventions to mitigate these disparities.

Homelessness continues to impact many individuals worldwide. It was estimated that at least 580,000 Americans experienced homelessness in 2018 (1). Of those experiencing homelessness, approximately 40% have at least one chronic medical condition (2). While all chronic conditions pose their own difficulties, diabetes necessitates rigorous self-monitoring, frequent medication dose adjustments, and stringent dietary adherence to prevent serious complications (3). Intensive self-management of diabetes presents a particular challenge for people with lived experience of homelessness (PWLEH) due to social and structural barriers (4) that may preclude access to medical care and healthy foods (5,6). While the prevalence of diabetes in PWLEH is at least as high as that in the general population (7), these barriers can contribute to suboptimal glycemia (8), which puts them at increased risk for complications.

Many studies highlight the link between socioeconomic status and adverse diabetes outcomes, such as suboptimal glycemia (9), microvascular complications (10), and hospitalizations for acute glycemic emergencies (11). Yet, relatively little is known about diabetes complications among a group of people who face an extreme socioeconomic disadvantage: those who are experiencing homelessness. It is known that PWLEH generally have increased rates of visits to the emergency department and higher rates of mortality (12,13). They are also known to be at increased risk for cardiovascular disease and amputations (14,15). However, many studies that focus on diabetes outcomes in this population use a qualitative methodology or small cohort studies reliant on self-reported outcomes. Therefore, our knowledge of the population-level impact of homelessness on diabetes complications is limited, which restricts our ability to quantify widespread disparities in outcomes and identify opportunities for interventions that improve resource utilization.

It can be challenging to identify PWLEH within population-level health data (16). This may be because privacy legislation in many jurisdictions necessitates housing and health data to be kept separate (17). However, in 2019, a Canadian algorithm for identifying homelessness in population-based administrative health data sources, using coded hospital information, was validated to ascertain homeless status at the time of acute care encounters (16). We sought to apply this algorithm to compare diabetes complications between patients with and without a history of homelessness at the population level to determine whether homelessness acts as an independent risk factor for adverse diabetes outcomes.

We conducted a propensity-matched cohort study using data from Ontario, Canada, at the ICES (formerly known as the Institute for Clinical Evaluative Sciences), an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement. This study was approved by the University of Calgary’s Conjoint Health Research Ethics Board (ID: REB20-1973). We followed the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) reporting guideline in the presentation of this research.

Administrative data were collected for all residents in the province of Ontario over the age of 18 years with a valid Ontario health card number. Eligible individuals required a diagnosis of diabetes, which was identified using a validated algorithm based on the following criteria: at least one hospital admission for diabetes or two outpatient physician claims for diabetes in a given year (18).

Next, we sought to identify homeless status using an administrative data algorithm, which was specifically validated in ICES data (16,19). These data sets were linked using unique encoded identifiers and analyzed at ICES. Since the algorithm for ascertaining homeless status is primarily based on acute health care encounters, we further restricted the cohort to individuals with diabetes who used acute care services for any reason during the accrual period. This was defined by at least one hospital admission or emergency department encounter from 1 April 2006 to 31 March 2019. These encounters were reported in the Ontario Mental Health Reporting System (OMHRS), which stores inpatient admissions to designated mental health beds; the National Ambulatory Care Reporting System (NACRS), which stores information on emergency department visits; or the Discharge Abstract Database (DAD), which stores information on admissions to acute medical facilities, rehabilitation centers, or psychiatric institutions. To be eligible for inclusion, at least one hospital encounter must have occurred after the criteria for diabetes and age were met.

The data elements and indicator codes used to identify homelessness are listed in Supplementary Table 1. Individuals were classified as homeless if they had any hospital encounter with documented homelessness in OMHRS, NACRS, or DAD during the study accrual period. This algorithm has high specificity (>99%), providing confidence that the people who are identified as homeless in administrative data are truly homeless at the time of the hospitalization.

For each person with a history of homelessness, an index date was assigned as the date of discharge from their first hospital encounter with documented homelessness that occurred during the accrual period and after all other criteria were met. As we could not use the same criteria for nonhomeless individuals (as no history of homelessness requires that no encounters have documented homelessness during the entire study period), we needed to approximate the same time from eligibility to index date in the homeless and nonhomeless groups. Therefore, for people with no history of homelessness, the index dates were assigned as the date of discharge from a randomly selected hospital encounter, based on the distribution of index dates in the group with a history of homelessness.

Eligible individuals with a history of homelessness were matched 1:1 to nonhomeless control subjects, based on propensity scores for relevant sociodemographic and clinical characteristics, as previously described (19). Sociodemographic characteristics included geographical region of residence. Given that people experiencing homelessness more often live in large urban centers, subjects were matched on region of residence to ensure comparison of individuals within similar regions and, therefore, similar access to care. Clinical characteristics included in the match were duration of diabetes, diabetes subtype, and past hospitalization for major adverse cardiovascular events, and mental illness or substance use were ascertained from hospital and physician claims databases.

Outcomes

Diabetes complications were ascertained from DAD and NACRS from 1 April 2006 to 31 March 2019, ensuring that each subject was followed for at least 1 year after their index date. Outcomes included macrovascular complications, revascularization procedures, and hospitalizations for acute glycemic emergencies, skin/soft tissue infections (SSTI), or amputation. Macrovascular complications were defined by hospital admission for cardiovascular events, such as myocardial infarction (MI), stroke, and heart failure (HF), as well as peripheral arterial disease presentations such as acute limb ischemia. Revascularization procedures were separated into coronary revascularization procedures such as percutaneous transluminal coronary angioplasty, including stenting (PTCA), or coronary artery bypass graft surgery (CABG), and peripheral arterial revascularization. Glycemic emergencies included hospital admission or emergency department visits for diabetic ketoacidosis (DKA), other hyperglycemic events (not DKA), and hypoglycemia. Finally, we assessed hospital admissions and emergency department visits for SSTI and amputations. The ICD-10 codes used to define these outcomes are reported in Supplementary Table 2.

Statistical Analysis

A logistic regression model was used to estimate the propensity score for homelessness, with homeless status as the dependent variable and sex, duration of diabetes (defined as time since entry into the Ontario Diabetes Dataset), type of diabetes, and past hospitalization for acute MI, HF, or stroke prior to index date as the independent variables. Patients with and without documented homelessness were then matched on index year, age within 5 years, a set of indicator variables for type of index hospital encounter, geographical region, presence of mental illness or substance use disorder, and the logit of the propensity score within 0.2 calipers using greedy matching without replacement. A comparison of the demographic and clinical characteristics for each group of people with and without documented homelessness was done using standardized differences, as previously described (19).

We calculated the rates of complications for people with documented homelessness and nonhomeless control subjects using offset terms to allow for different durations of time at risk. Rate ratios (RRs) for diabetes complications and revascularization procedures were calculated using generalized linear models with negative binomial distribution and robust SEs (20).

Subsequently, a set of sensitivity analyses were completed. First, it can be challenging to separate the impact of homelessness from other social determinants such as socioeconomic status. To ensure that low income was not driving our results, a second cohort was created, matching each person with a history of homelessness to nonhomeless control subjects living in the lowest neighborhood income quintile (21). Next, the initial match criteria included any diagnosis of mental illness but did not specify type of mental illness. While we expect the severity of illness to be relatively balanced, due to inclusion of a variable for hospitalization to a designated psychiatric inpatient bed (OMHRS), we recognize that certain medications for schizophrenia (e.g., second-generation antipsychotics) can impact cardiometabolic disease (22). To ensure type of mental illness did not impact our results, we conducted a sensitivity analysis that included a variable for type of mental illness in the match criteria.

All analyses were conducted using SAS 7.1 software (SAS Institute, Toronto, Ontario, Canada).

Patient Characteristics

Overall, 1,076,437 eligible people with diabetes were identified during the study period. Of these, 6,944 had a history of homelessness. A suitable nonhomeless match was found for 5,219 people with documented homelessness (75%). Baseline demographic and clinical characteristics are summarized in Table 1. The mean age of the matched cohort was 55 years. There were more men (67.7% and 66.7% for those with a history of homelessness vs. nonhomeless control subjects, respectively) than women (32.3% vs. 33.3%). A large proportion of the cohort was from a large urban area (Central Toronto, 27.2% for both cohorts), and most had a mental illness (74.7% for both cohorts). Approximately one-third of the cohort had diabetes for >20 years (36.8% vs. 37.1%), and the vast majority were considered to have type 2 diabetes (97.6% vs. 97.9%).

Table 1

Baseline demographic and clinic characteristics for the matched (N = 10,438) cohorts comparing demographic and clinical characteristics of patients with a history of homelessness and nonhomeless control subjects

Demographic and clinical characteristicsHomeless (n = 5,219)Nonhomeless control subjects (n = 5,219)Standardized difference
Age at index, years    
 Mean ± SD 55.20 ± 13.58 55.57 ± 13.69 0.027 
 Median (interquartile range) 55 (46–64) 55 (46–64) 0.026 
Age-group    
 18–44 years 1,142 (21.9) 1,142 (21.9) 
 45–64 years 2,837 (54.4) 2,837 (54.4)  
 ≥65 years 1,240 (23.8) 1,240 (23.8)  
Sex    
 Female 1,688 (32.3) 1,737 (33.3) 0.02 
 Male 3,531 (67.7) 3,482 (66.7)  
Location (Local Health Integration Network)    
 Toronto Central 1,420 (27.2) 1,420 (27.2) 
 Central East 547 (10.5) 547 (10.5)  
 Hamilton, Niagara, Haldimand, Brant 514 (9.8) 514 (9.8)  
 Southwest 448 (8.6) 448 (8.6)  
 Champlain 430 (8.2) 430 (8.2)  
 Central 334 (6.4) 334 (6.4)  
 Other 1,526 (29.2) 1,526 (29.2)  
Hospital encounter type (index date ± 6 months)    
 Psychiatric inpatient admission to mental health bed (OMHRS) 1,254 (24.0) 1,254 (24.0) 
 Other inpatient admission (DAD) 1,149 (22.0) 1,149 (22.0) 
 Emergency department (NACRS) 4,911 (94.1) 4,911 (94.1) 
Chronic conditions    
 Any mental illness 3,900 (74.7) 3,900 (74.7) 
 Acute MI 604 (11.6) 646 (12.4) 0.025 
 HF 699 (13.4) 838 (16.1) 0.075 
 Stroke 426 (8.2) 552 (10.6) 0.083 
Duration of diabetes    
 <5 years 1,051 (20.1) 1,029 (19.7) 0.011 
 5–10 years 1,040 (19.9) 1,043 (20.0) 0.001 
 10–20 years 1,209 (23.2) 1,213 (23.2) 0.002 
 ≥20 years 1,919 (36.8) 1,934 (37.1) 0.006 
Type of diabetes    
 Type 1 diabetes 127 (2.4) 109 (2.1) 0.023 
 Type 2 diabetes 5,092 (97.6) 5,110 (97.9)  
Demographic and clinical characteristicsHomeless (n = 5,219)Nonhomeless control subjects (n = 5,219)Standardized difference
Age at index, years    
 Mean ± SD 55.20 ± 13.58 55.57 ± 13.69 0.027 
 Median (interquartile range) 55 (46–64) 55 (46–64) 0.026 
Age-group    
 18–44 years 1,142 (21.9) 1,142 (21.9) 
 45–64 years 2,837 (54.4) 2,837 (54.4)  
 ≥65 years 1,240 (23.8) 1,240 (23.8)  
Sex    
 Female 1,688 (32.3) 1,737 (33.3) 0.02 
 Male 3,531 (67.7) 3,482 (66.7)  
Location (Local Health Integration Network)    
 Toronto Central 1,420 (27.2) 1,420 (27.2) 
 Central East 547 (10.5) 547 (10.5)  
 Hamilton, Niagara, Haldimand, Brant 514 (9.8) 514 (9.8)  
 Southwest 448 (8.6) 448 (8.6)  
 Champlain 430 (8.2) 430 (8.2)  
 Central 334 (6.4) 334 (6.4)  
 Other 1,526 (29.2) 1,526 (29.2)  
Hospital encounter type (index date ± 6 months)    
 Psychiatric inpatient admission to mental health bed (OMHRS) 1,254 (24.0) 1,254 (24.0) 
 Other inpatient admission (DAD) 1,149 (22.0) 1,149 (22.0) 
 Emergency department (NACRS) 4,911 (94.1) 4,911 (94.1) 
Chronic conditions    
 Any mental illness 3,900 (74.7) 3,900 (74.7) 
 Acute MI 604 (11.6) 646 (12.4) 0.025 
 HF 699 (13.4) 838 (16.1) 0.075 
 Stroke 426 (8.2) 552 (10.6) 0.083 
Duration of diabetes    
 <5 years 1,051 (20.1) 1,029 (19.7) 0.011 
 5–10 years 1,040 (19.9) 1,043 (20.0) 0.001 
 10–20 years 1,209 (23.2) 1,213 (23.2) 0.002 
 ≥20 years 1,919 (36.8) 1,934 (37.1) 0.006 
Type of diabetes    
 Type 1 diabetes 127 (2.4) 109 (2.1) 0.023 
 Type 2 diabetes 5,092 (97.6) 5,110 (97.9)  

Data are presented as n (%) unless indicated otherwise.

There was no significant difference in the baseline rate of diabetes complications prior to cohort entry into the study between matched groups. However, prior admissions for acute MI, HF, or stroke were slightly higher in the nonhomeless control group (Table 1).

Our matched low-income comparison group was slightly smaller due to finding fewer suitable matches in a smaller comparison population pool, with 4,670 matched pairs. However, it was otherwise similar in most respects. The clinical and demographic characteristics for both groups of this match are provided in Supplementary Table 3.

Macrovascular Complications and Revascularization Procedures

The rate of macrovascular complications among people with a history of homelessness was 137.3 per 1,000 person-years, which was considerably higher than the rate of 72.4 per 1,000 person-years among those without a history of homelessness (RR 1.85, 95% CI 1.64–2.07) (Fig. 1 and Table 2). Specifically, their rates of complications were higher for acute MI (RR 1.26, 95% CI 1.02–1.54), stroke (RR 1.41, 95% CI 1.15–1.73), HF (RR 1.78, 95% CI 1.51–2.11), and peripheral artery disease (RR 2.92, 95% CI 2.38–3.60) compared with nonhomeless control subjects (Fig. 1).

Figure 1

RRs of outcomes of interest, for people with experience of homelessness vs. those with no experience of homelessness. A: Macrovascular complications. B: Revascularization procedures. C: Hospital visits for other diabetes complications.

Figure 1

RRs of outcomes of interest, for people with experience of homelessness vs. those with no experience of homelessness. A: Macrovascular complications. B: Revascularization procedures. C: Hospital visits for other diabetes complications.

Close modal
Table 2

Rates of macrovascular complications, revascularization procedures, and other diabetes complications among people with diabetes with a history of homelessness compared with nonhomeless control subjects

Diabetes complications and proceduresHistory of homelessness(n = 5,219)Nonhomeless control subjects(n = 5,219)RR (95% CI)
Events(n)Rate of events per 1,000 person-yearsEvents(n)Rate of events per 1,000 person-yearsHomeless vs. nonhomeless
Macrovascular complications 
 Acute MI 466 18.7 436 16.8 1.26 (1.02, 1.54) 
 Stroke 278 11.2 223 8.6 1.41 (1.15, 1.73) 
 HF 1,441 57.9 772 29.7 1.78 (1.51, 2.11) 
 Peripheral arterial disease 1,229 49.4 449 17.3 2.92 (2.38, 3.60) 
 Macrovascular composite 3,414 137.3 1,880 72.4 1.85 (1.64, 2.07) 
Revascularization procedures 
 Coronary revascularizationa 251 10.1 357 13.8 0.76 (0.62, 0.92) 
 CABG 55 2.2 90 3.5 0.63 (0.45, 0.89) 
 PCTA 196 7.9 267 10.3 0.79 (0.62, 1.00) 
 Peripheral revascularization 120 4.8 92 3.5 1.41 (0.96, 2.06) 
Other diabetes-related complications 
 DKA 1,001 40.3 213 8.2 5.64 (4.07, 7.81) 
 Hyperglycemia (no DKA) 134 5.4 53 2.72 (1.90, 3.90) 
 Hypoglycemia 1,188 47.8 452 17.4 2.62 (2.02, 3.39) 
 SSTI 4,958 199.4 1,384 53.3 3.78 (3.31, 4.32) 
 Amputations 168 6.8 88 3.4 2.32 (1.65, 3.27) 
Diabetes complications and proceduresHistory of homelessness(n = 5,219)Nonhomeless control subjects(n = 5,219)RR (95% CI)
Events(n)Rate of events per 1,000 person-yearsEvents(n)Rate of events per 1,000 person-yearsHomeless vs. nonhomeless
Macrovascular complications 
 Acute MI 466 18.7 436 16.8 1.26 (1.02, 1.54) 
 Stroke 278 11.2 223 8.6 1.41 (1.15, 1.73) 
 HF 1,441 57.9 772 29.7 1.78 (1.51, 2.11) 
 Peripheral arterial disease 1,229 49.4 449 17.3 2.92 (2.38, 3.60) 
 Macrovascular composite 3,414 137.3 1,880 72.4 1.85 (1.64, 2.07) 
Revascularization procedures 
 Coronary revascularizationa 251 10.1 357 13.8 0.76 (0.62, 0.92) 
 CABG 55 2.2 90 3.5 0.63 (0.45, 0.89) 
 PCTA 196 7.9 267 10.3 0.79 (0.62, 1.00) 
 Peripheral revascularization 120 4.8 92 3.5 1.41 (0.96, 2.06) 
Other diabetes-related complications 
 DKA 1,001 40.3 213 8.2 5.64 (4.07, 7.81) 
 Hyperglycemia (no DKA) 134 5.4 53 2.72 (1.90, 3.90) 
 Hypoglycemia 1,188 47.8 452 17.4 2.62 (2.02, 3.39) 
 SSTI 4,958 199.4 1,384 53.3 3.78 (3.31, 4.32) 
 Amputations 168 6.8 88 3.4 2.32 (1.65, 3.27) 
a

No endarterectomy procedures were completed.

In contrast, the rate of coronary revascularization procedures for those with a history of homelessness was 10.1 per 1,000 person-years, which was lower than the 13.8 per 1,000 person-years for people without a history of homelessness (RR 0.76, 95% CI 0.62–0.94). The rate of CABG procedures was 37% lower (RR 0.63, 95% CI 0.45–0.89) and the rate of PTCA was 21% lower (RR 0.79, 95% CI 0.62–1.00) than those with no history of homelessness. However, the rate of peripheral revascularization procedures was not statistically different between the groups, with a possible trend toward a higher rate among those with a history of homelessness (RR 1.41, 95% CI 0.96–2.06) (Fig. 1).

When the low-income match was used, similar results were observed, with even stronger RRs reported for some complications. For example, people with documented homelessness had 2.20 times higher rate of macrovascular complications than low-income control subjects (95% CI 1.86–2.38) (Fig. 1). The most notable difference was for peripheral artery disease, with a RR of 3.50 (95% CI 2.77–4.42). Similar to the overall match, the rate of coronary revascularization procedures was lower, but the rate of peripheral revascularization was slightly although statistically significantly higher among people with documented homelessness (Supplementary Table 4).

In the revised match incorporating type/severity of mental illness, the same patterns were seen with higher rates of macrovascular complications (RR 1.55, 95% CI 1.36–1.77). However, differences in rates of coronary revascularization were attenuated (RR 1.05, 95% CI 0.84–1.31) (Supplementary Table 5).

Hospitalizations for Glycemia

The rate of DKA presentations in people with a history of homelessness was 40.3 per 1,000 person-years compared with 8.2 per 1,000 person-years among those with no history of homelessness (RR 5.64; 95% CI 4.07–7.81). The rate of hospital visits for non-DKA hyperglycemia was 2.72 times that of those with no history of homelessness (95% CI 1.90–3.90), while the rate of hypoglycemia-related presentations was 2.62 times that of the control group (95% CI 2.02–3.39). The same patterns were seen for glycemia-related hospitalizations when the group with a history of homelessness was compared with the low-income control subjects (Supplementary Table 4) and when the match was revised for type/severity of mental illness (Supplementary Table 5).

SSTIs and Amputations

Those who had a history of homelessness were also more likely to have SSTI (199.4 per 1,000 person-years vs. 53.3 per 1,000 person-years), with a RR of 3.78 (95% CI 3.31–4.32). The rate of amputations was 2.32 times the rate of the nonhomeless control group (95% CI 1.65–3.27). These same trends were also observed in the low-income match and when matching for type of mental illness.

Our findings show that people with a history of homelessness have substantially higher rates of diabetes complications compared with matched individuals without a history of homelessness. Specifically, rates of cardiovascular complications, including MI, stroke, HF, and peripheral arterial disease, were higher among people with a history of homelessness than nonhomeless control subjects. Further, these rates of complications remained higher among people with a history of homelessness even in the low income match. Thus, even though low neighborhood income quintile is associated with poor glycemic control (9), this factor does not solely account for the disparity in outcomes observed. Overall, this suggests that a history of homelessness may independently increase risk for diabetes complications above and beyond that observed related to income disparities. This study adds to the understanding of our previously reported finding of increased mortality among those with a history of homelessness (23). Furthermore, since the match criteria balanced other factors, such as mental illness and region of residence, the differences in diabetes outcomes seen in this population cannot be solely attributed to mental health status or geographical disparities in access to care.

While our study is the first to examine these associations within a cohort of people with diabetes, these results are consistent with previous research that links homelessness with incidence of cardiovascular disease (24). It should be noted that PWLEH are known to have higher rates of smoking, which itself is a strong risk factor for cardiovascular disease (7). We expect that smoking contributed to the causal pathway between homelessness and cardiovascular outcomes among people with diabetes. As such, we would not have included smoking in our match criteria; however, having the data to assess the differences in prevalence of smoking across groups would be beneficial. This could be a potential area for future work to optimize diabetes outcomes in PWLEH.

Further, people with diabetes and a history of homelessness often report considerable difficulties with self-management behaviors and report worse overall general health status (25,26). Inability to access healthy foods or take medications at regular times greatly impacts this population’s ability to maintain normoglycemia, which may lead to chronic hyperglycemia and increased risk for macrovascular complications (4,12). Our findings demonstrate that rates of acute glycemic emergencies were higher in those with a history of homelessness. Homelessness has been shown to be associated with increased hospital admissions and readmission for DKA (27). Repeated presentation suggests considerable difficulty maintaining normoglycemia, which can also be due to an inability to store or access insulin or medications when needed (28).

Despite the higher rate of MI in PWLEH, we observed that the rate of coronary revascularization procedures was considerably lower among people with documented homelessness compared with the nonhomeless group. Eligibility for revascularization depends on the time of presentation with acute MI, cardiac viability, anatomy of vasculature, and previous revascularization procedures performed (29). Thus, there are many factors that could have affected whether revascularization procedures were offered or performed. Given the reported difficulty with accessing outpatient care and follow-up for PWLEH (30), the lower rate of procedures also raises the question of whether PWLEH have later and more severe presentations of MIs, making them less likely to be eligible for reperfusion. One small study in Toronto noted that people with and without a history of homelessness had comparable rates of cardiac events, but they also found lower rates of PTCA and higher rates of cardiac mortality in their homeless cohort. Reasons for fewer interventions in this population included concern for poor adherence to antiplatelet therapy that is required after stenting (31).

Unlike coronary revascularization, the rate of peripheral revascularization procedures was slightly higher in the group with documented homelessness than in the nonhomeless and low-income control groups. However, despite the higher rate of peripheral revascularization, the difference in rates was still much lower than the disparity in peripheral artery disease and SSTI, suggesting that the rate may still not reflect the greater need for intervention in the population. Further, amputations were more frequent than peripheral revascularization, raising the question of the pathways PWLEH have to access care and follow-up for peripheral artery disease. The propensity for amputation over revascularization could be suggestive of more severe initial presentations of peripheral artery disease or higher rates of concomitant SSTI. Indeed, our results and those of other studies support higher SSTI rates among those with a history of homelessness (32).

Although our study offers novel insights of the effects of homelessness on diabetes outcomes, it is also important to consider its limitations.

First, the cohort was limited to people who accessed hospital services. Therefore, the cohort did not include people who only presented to outpatient clinics or who never sought medical care. Thus, we can only generalize these data to people who use acute care services.

Second, we were only able to identify homeless status at hospital encounter based on the diagnostic codes. The algorithm had low sensitivity, which means not all people who were homeless could be identified using hospital encounters. People with more frequent hospital encounters are more likely to be coded and captured as homeless, which could lead to selection bias in our homeless cohort. It is also possible that some people with identified homelessness could have become housed over the study period, while others identified as nonhomeless could have experienced homelessness between hospital encounters. Again, this could not be captured by our algorithm; however, if the groups became more similar to one another with respect to their homeless status over time, we expect that this would, in fact, bias our results toward the null. Further, among the data elements used to identify homelessness in the validated algorithm, there was one code for “supportive housing,” which may carry a different risk than living in a shelter or on the street. However, only 12 individuals were enrolled in our cohort based on a supportive housing code (0.2% of matches), so this likely had minimal impact on our results.

Third, second-generation antipsychotics are known to increase risk for cardiometabolic disease, and they may have acted as confounders between homelessness and cardiovascular outcomes, specifically. To partially address this as a potential confounder, we performed a sensitivity analysis by including type/severity of mental illness in our match criteria. The findings were largely unchanged, although somewhat attenuated.

Finally, our propensity score was limited by the information available in the ICD-10 codes, and as such, we could not control for confounders such as ethnicity, which can influence global cardiovascular risk and may account for some of the differences seen in our results. Previous studies have shown among PWLEH in Toronto, 60% identify as a racialized group, with ∼31% identifying as Black, 10% as Indigenous, and 10% as Asian (33). Although these data cannot be directly applied to our own study, they can be used for context when determining whether the data from our study population may apply to other large cities in Canada.

Diabetes is a chronic condition that places a high burden of responsibility on patients. The disease necessitates daily monitoring, dietary adherence, and medication adjustment to prevent hyperglycemia, which can lead to both dangerous acute events and long-term macrovascular complications (34). PWLEH are known to be at increased risk for hyperglycemia (8), but up until recently, the ultimate results had not been studied at a population level. Using a propensity-matched cohort design, we have shown that those with a history of homelessness have increased rates of both acute and chronic macrovascular complications of diabetes. This provides us with the groundwork to further study how unstable housing can lead to disparities in diabetes outcomes. Some of these disparities might arise from unmet health care needs such as access to primary care, elective medical procedures, prescription medications, as well as basic needs such as access to healthy food and clean living quarters (3537).

While it is already known that unstable housing is associated with poorer health outcomes, generally (13,36), our study identifies homelessness as an independent risk factor for diabetes complications. Although our study was limited to macrovascular complications and acute hyperglycemia, microvascular complications are also known to present a high burden of disease among these patients and is an important area of further research. Similarly, while our study focused on patients who presented to inpatient care, it would be interesting to investigate whether the disparities seen persist in outpatient care as well. With increased burden of diabetes complications on people experiencing homelessness, funding for targeted outreach practices to improve health outcomes in this population is essential (38). Encouraging interventions such as outpatient screening teams, smoking cessation programs, shelter-based monitoring and follow-up programs (39), and housing supports (37) may decrease preventable complications among PWLEH.

R.S. and K.W. share first authorship.

This article contains supplementary material online at https://doi.org/10.2337/figshare.22816055.

Funding. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC). This study also received funding from MSI Foundation Grant to D.J.T.C. and a Diabetes Action Canada Early Career Mentorship Award to D.J.T.C, P.E.R., and G.L.B. This document used data adapted from the Statistics Canada Postal CodeOM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from Canada Post Corporation and Statistics Canada. Parts of this material are based on data and/or information compiled and provided by Canadian Institute for Health Information.

The analyses, conclusions, opinions and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. No endorsement by the OHDP, its partners, or the Province of Ontario is intended or should be inferred.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.S., K.W., L.B., and D.J.T.C. researched data and wrote, reviewed, and edited the manuscript. P.E.R., S.W.H., G.L.B., P.C.A., and E.S. reviewed and edited the manuscript. All authors approved the final version of the manuscript. D.J.T.C. 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.

Prior Presentation. This work was presented as an oral presentation at the Canadian Association of Health Services and Policy Research, virtual meeting, 31 May–2 June 2022.

1.
The White House
.
The State of Homelessness in America
.
The Council of Economic Advisors, Executive Summary. 2019. Accessed 6 February 2023. Available from https://trumpwhitehouse.archives.gov/wp-content/uploads/2019/09/The-State-of-Homelessness-in-America.pdf
2.
Plumb
JD
.
Homelessness: care, prevention, and public policy
.
Ann Intern Med
1997
;
126
:
973
975
3.
Clement
M
,
Harvey
B
,
Rabi
DM
,
Roscoe
RS
;
Canadian Diabetes Association Clinical Practice Guidelines Expert Committee
.
Organization of diabetes care
.
Can J Diabetes
2013
;
37
(
Suppl. 1
):
S20
S25
4.
Grewal
EK
,
Campbell
RB
,
Booth
GL
, et al
.
Using concept mapping to prioritize barriers to diabetes care and self-management for those who experience homelessness
.
Int J Equity Health
2021
;
20
:
158
5.
Campbell
DJ
,
O’Neill
BG
,
Gibson
K
,
Thurston
WE
.
Primary healthcare needs and barriers to care among Calgary’s homeless populations
.
BMC Fam Pract
2015
;
16
:
139
6.
Campbell
RB
,
Larsen
M
,
DiGiandomenico
A
, et al
.
The challenges of managing diabetes while homeless: a qualitative study using photovoice methodology
.
CMAJ
2021
;
193
:
E1034
E1041
7.
Lee
TC
,
Hanlon
JG
,
Ben-David
J
, et al
.
Risk factors for cardiovascular disease in homeless adults
.
Circulation
2005
;
111
:
2629
2635
8.
Axon
RN
,
Gebregziabher
M
,
Dismuke
CE
, et al
.
Differential impact of homelessness on glycemic control in veterans with type 2 diabetes mellitus
.
J Gen Intern Med
2016
;
31
:
1331
1337
9.
McBrien
KA
,
Manns
BJ
,
Hemmelgarn
BR
, et al
.
The association between sociodemographic and clinical characteristics and poor glycaemic control: a longitudinal cohort study
.
Diabet Med
2016
;
33
:
1499
1507
10.
Bachmann
MO
,
Eachus
J
,
Hopper
CD
, et al
.
Socio-economic inequalities in diabetes complications, control, attitudes and health service use: a cross-sectional study
.
Diabet Med
2003
;
20
:
921
929
11.
Booth
GL
,
Hux
JE
.
Relationship between avoidable hospitalizations for diabetes mellitus and income level
.
Arch Intern Med
2003
;
163
:
101
106
12.
Lin
WC
,
Bharel
M
,
Zhang
J
,
O’Connell
E
,
Clark
RE
.
Frequent emergency department visits and hospitalizations among homeless people with Medicaid: implications for Medicaid expansion
.
Am J Public Health
2015
;
105
(
Suppl. 5
):
S716
S722
13.
Cheung
AM
,
Hwang
SW
.
Risk of death among homeless women: a cohort study and review of the literature
.
CMAJ
2004
;
170
:
1243
1247
14.
Baggett
TP
,
Liauw
SS
,
Hwang
SW
.
Cardiovascular disease and homelessness
.
J Am Coll Cardiol
2018
;
71
:
2585
2597
15.
To
MJ
,
Brothers
TD
,
Van Zoost
C
.
Foot conditions among homeless persons: a systematic review
.
PLoS One
2016
;
11
:
e0167463
16.
Richard
L
,
Hwang
SW
,
Forchuk
C
, et al
.
Validation study of health administrative data algorithms to identify individuals experiencing homelessness and estimate population prevalence of homelessness in Ontario, Canada
.
BMJ Open
2019
;
9
:
e030221
17.
Thomas
I
,
Tweed
E
.
The promises and pitfalls of administrative data linkage for tackling homelessness
.
Eur J Homelessness
2021
;
15
:
177
188
18.
Hux
JE
,
Ivis
F
,
Flintoft
V
,
Bica
A
.
Diabetes in Ontario: determination of prevalence and incidence using a validated administrative data algorithm
.
Diabetes Care
2002
;
25
:
512
516
19.
Wiens
K
,
Bai
L
,
Austin
PC
, et al
.
Characteristics of people with type I or type II diabetes with and without a history of homelessness: a population-based cohort study
.
1 January 2022 [preprint]. medRxiv 2022.08.11.22278127
20.
Weaver
CG
,
Ravani
P
,
Oliver
MJ
,
Austin
PC
,
Quinn
RR
.
Analyzing hospitalization data: potential limitations of Poisson regression
.
Nephrol Dial Transplant
2015
;
30
:
1244
1249
21.
Austin
PC
.
Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples
.
Stat Med
2009
;
28
:
3083
3107
22.
American Diabetes Association
;
American Psychiatric Association
;
American Association of Clinical Endocrinologists
;
North American Association for the Study of Obesity
.
Consensus development conference on antipsychotic drugs and obesity and diabetes
.
Diabetes Care
2004
;
27
:
596
601
23.
Wiens
K
,
Bai
L
,
Austin
PC
, et al
.
Long-term association between homelessness and mortality among people with diabetes
.
Lancet Diabetes Endocrinol
2023
;
11
:
229
231
24.
Nanjo
A
,
Evans
H
,
Direk
K
,
Hayward
AC
,
Story
A
,
Banerjee
A
.
Prevalence, incidence, and outcomes across cardiovascular diseases in homeless individuals using national linked electronic health records
.
Eur Heart J
2020
;
41
:
4011
4020
25.
Keene
DE
,
Guo
M
,
Murillo
S
.
“That wasn’t really a place to worry about diabetes”: housing access and diabetes self-management among low-income adults
.
Soc Sci Med
2018
;
197
:
71
77
26.
Mosley-Johnson
E
,
Walker
RJ
,
Nagavally
S
, et al
.
Relationship between food insecurity and housing instability on quality of care and quality of life in adults with diabetes
.
PLoS One
2022
;
17
:
e0278650
27.
Sakai-Bizmark
R
,
Mena
LA
,
Estevez
D
, et al
.
Health care utilization of homeless minors with diabetes in New York state from 2009 to 2014
.
Diabetes Care
2020
;
43
:
2082
2089
28.
Edidin
JP
,
Ganim
Z
,
Hunter
SJ
,
Karnik
NS
.
The mental and physical health of homeless youth: a literature review
.
Child Psychiatry Hum Dev
2012
;
43
:
354
375
29.
O’Gara
PT
,
Kushner
FG
,
Ascheim
DD
, et al.;
CF/AHA Task Force
.
2013 ACCF/AHA guideline for the management of ST-elevation myocardial infarction: executive summary: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines
.
Circulation
2013
;
127
:
529
555
30.
Corrigan
P
,
Pickett
S
,
Kraus
D
,
Burks
R
,
Schmidt
A
.
Community-based participatory research examining the health care needs of African Americans who are homeless with mental illness
.
J Health Care Poor Underserved
2015
;
26
:
119
133
31.
Liauw
SSL
,
Luong
L
,
Liu
S
, et al
.
Clinical presentation and outcome of patients experiencing homelessness presenting with ST-segment elevation myocardial infarction
.
Can J Cardiol
2021
;
37
:
1555
1561
32.
Lewis
SD
,
Peter
GS
,
Gómez-Marín
O
,
Bisno
AL
.
Risk factors for recurrent lower extremity cellulitis in a U.S. Veterans Medical Center population
.
Am J Med Sci
2006
;
332
:
304
307
33.
City of Toronto
.
Street Needs Assessment 2021: Attachment 1
.
City of Toronto, 2022. Accessed 3 March 2023. Available from: https//www.toronto.ca/wp-content/uploads/2022/11/96bf-SSHA-2021-Street-Needs-Assessment.pdf
34.
Heinrich
E
,
Schaper
NC
,
de Vries
NK
.
Self-management interventions for type 2 diabetes: a systematic review
.
Eur Diabetes Nurs
2010
;
7
:
71
76
35.
Hwang
SW
,
Ueng
JJ
,
Chiu
S
, et al
.
Universal health insurance and health care access for homeless persons
.
Am J Public Health
2010
;
100
:
1454
1461
36.
Baggett
TP
,
O’Connell
JJ
,
Singer
DE
,
Rigotti
NA
.
The unmet health care needs of homeless adults: a national study
.
Am J Public Health
2010
;
100
:
1326
1333
37.
Omerov
P
,
Craftman
ÅG
,
Mattsson
E
,
Klarare
A
.
Homeless persons’ experiences of health- and social care: A systematic integrative review
.
Health Soc Care Community
2020
;
28
:
1
11
38.
Mancini
N
,
Campbell
R
,
Yaphe
H
, et al
.
Identifying challenges and solutions to providing diabetes care for those experiencing homelessness
.
Int J Homelessness
.
2022
;
2
:
48
67
39.
Scott
S
,
Grewal
EK
,
Tariq
H
,
Campbell
DJ
.
Developing and evaluating a SAFER model to screen for diabetes complications among people experiencing homelessness: a pilot study protocol
.
Pilot Feasibility Stud
2022
;
8
:
211
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.