Introduction

Diabetes is now the eighth leading, and most rapidly growing, cause of the global disease burden [1, 2]. Diabetes-related foot disease (DFD) is the leading cause of the diabetes disability burden [3,4,5]. DFD is defined by ulceration, infection or other destructive conditions of the foot in people with diabetes and is precipitated by the risk factors peripheral neuropathy and peripheral artery disease (PAD) [6, 7]. DFD takes months to heal, is intensive to treat and places people at high risk of hospital admission and amputation [8,9,10,11]. It has been estimated that 20 million people worldwide have DFD, with a further 130 million having a key risk factor for DFD, resulting in around nine million hospital admissions and 2 million amputations each year [3,4,5, 9].

DFD has long been known to be the leading cause of amputations [3,4,5]. For this reason, and because data on amputation procedures are relatively easily and reliably collected [12,13,14], the incidence of hospital admissions in which an amputation procedure has occurred has been used as the primary measure to monitor DFD burden [15,16,17]. Previous reviews have suggested that trends in the incidence of admissions for diabetes-related major amputations (above ankle) are decreasing over time, whereas admissions for minor amputations (below ankle) are increasing over time [15,16,17]. These previous reviews infer that DFD care must be improving, with hospital treatments such as minor amputation procedures more commonly achieving limb salvage and reductions in major amputation admissions [12,13,14]. However, these reviews have investigated admissions for amputations only [15, 16] and not considered any of the multiple other DFD conditions such as ulceration or infection. Furthermore, the last review to primarily investigate global amputation trends was published over a decade ago [17].

Amputations are now performed in a minority of people admitted to hospital for treatment of DFD, because other treatments such as antibiotics and revascularisation are being used more effectively to salvage limbs [5, 9, 18, 19]. However, DFD has recently been reported as a leading and growing cause of all hospital admissions in high-income countries [8,9,10, 18]. As most hospital admissions for DFD are now unrelated to amputation, the burden of DFD may be better assessed through hospital admission rates for all DFD conditions, in line with methods used to monitor other diseases [20,21,22,23]. The primary aim of this review was to synthesise the findings from peer-reviewed publications reporting the incidence of hospital admissions for DFD conditions or for diabetes-related amputations in nationally representative populations. Secondary aims included reviewing the incidence of or trends in hospital admissions for different classification groups, including DFD condition, amputation procedure, age, sex, diabetes type and country.

Methods

For this narrative review, we systematically searched PubMed and Embase to identify peer-reviewed publications reporting the incidence, or trends in the incidence, of hospital admissions for DFD conditions or diabetes-related amputation procedures in nationally representative populations between 1 January 2001 and 5 May 2022. This period was selected as it followed the publication of the first practice guidelines by the International Working Group on the Diabetic Foot (IWGDF) in 1996, which is likely to have substantially changed international practice [19]. We used search strings combining keywords for DFD conditions OR amputation procedures, incidence OR trend AND nation OR country (electronic supplementary material [ESM] Tables 1 and 2), which had been shown to identify ten key eligible publications [3, 23,24,25,26,27,28,29,30,31]. This review was registered in the Open Science Framework database (https://doi.org/10.17605/OSF.IO/4TZFJ).

Eligible publications were those reporting the incidence of hospital admissions for DFD conditions or diabetes-related amputations among a population considered representative of a country. Hospital admission for a DFD condition was defined as admission of a person with diabetes as an inpatient with a principal (primary reason for the admission) or additional (secondary reason for the admission) diagnosis of a DFD condition [9, 32]. A DFD condition was defined as ulceration, infection, PAD or neuropathy of the foot or lower extremity in a person with diabetes [6, 7]. An amputation was defined as any resection of a lower extremity through a bone or joint during a hospital admission of a person with diabetes. Amputations were categorised as major (proximal to the ankle) or minor (through or distal to the ankle) [6, 7]. Incidence rates in this review are reported for principal or any (principal or additional diagnosis) admissions for total DFD conditions combined (where publications reported aggregated admissions for DFD conditions, including ulceration and infection as a minimum) or for ulceration, infection, PAD or neuropathy separately (where publications reported these conditions separately). Incidence rates are also reported for hospital admissions for all amputations combined (where publications reported aggregated admissions for major and minor amputations) and for major and minor amputations separately (when reported separately). In this review, unless otherwise indicated, reported outcomes refer only to the primary DFD condition or highest-level amputation procedure that was investigated for each admission. We considered a nationally representative population as one that captures outcomes for at least 50% of the country’s population with diabetes, or at least 50% of a country’s resident population (i.e. the total population of all citizens of the country, with or without diabetes), or as one defined as a nationally representative sample according to the publication [16, 33]. Countries were categorised according to the World Bank income classification into high-, middle- or low-income countries [34].

Titles and abstracts identified in the searches were screened using the eligibility criteria by one author (PAL) and the full texts of eligible publications were retrieved and assessed for final inclusion by the same author. Relevant reviews identified in the searches and reference lists of included publications were also manually searched to identify additional eligible publications. Data were extracted from included publications into evidence tables by one author (PAL) and samples of extracted data were serially checked by another author (SMC) until no errors were found. Data extracted included country, study reference, aims, years of data collected, data sources, type of population, type of incidence, initial rate (the first incidence rate for the time period reported by the publication), end rate (the last incidence rate for the time period reported by the publication), relative trend over time (calculated descriptively as the percentage change between the initial rate and the end rate for the period reported) and p value if reported. Publications reporting similarly defined incidence rates were descriptively summarised using medians (ranges: min. value, max. value reported) for initial rates, end rates and relative trends. For ease of reporting results, we descriptively defined a decreasing trend as a >5% decrease in relative trend, an increasing trend as a >5% increase in relative trend and a stable trend as a ≤5% relative trend in either direction; however, these rates do not infer statistical significance.

Results

A total of 3990 records were identified as relevant from the initial searches, resulting in 2527 unique records after duplicate removal. After title and abstract screening, 117 publications were deemed eligible for full-text assessment, with 57 ultimately included after full-text review. Further manual searching identified another 14 publications for inclusion, leading to 71 publications being included in this review. Of these 71 publications, 14 (20%) reported on admissions for DFD conditions and 66 (93%) on admissions for amputations (nine [13%] reported both). Table 1 summarises the main findings of these publications.

Table 1 Summary incidence of hospital admissions for DFD conditions and amputations per 1000 person-years with diabetes
Full size table

DFD conditions

We found 14 publications reporting the incidence rates of hospital admissions for DFD conditions (Table 2, ESM Table 3). Four publications reported incidence rates for total DFD conditions combined, seven reported rates for ulceration, three reported rates for infection and four reported rates for PAD. No publications reported incidence rates for neuropathy. Of the 14 publications, 13 (93%) reported rates using the population with diabetes as a denominator and two (14%) reported rates among the resident population (one [7%] reported both). All reported rates for high-income countries, including three Asian countries, two European countries, two North American countries and one Western Pacific country. Population denominators were identified from national health insurance databases (57%), national diabetes registries (21%), national diabetes prevalence surveys applied to national census databases (14%) and national census databases (7%). All publications identified incidence of admission for DFD condition numerators using ICD diagnosis codes (50% used ICD-9 [http://www.icd9data.com/2007/Volume1/default.htm] and 50% used ICD-10 [http://apps.who.int/classifications/icd10/browse/2016/en] diagnosis codes) from national health insurance databases (57%), national hospital discharge databases (36%) or national inpatient samples (7%). All publications reported a mutually exclusive incidence of admission for a DFD condition, with 36% of publications reporting only principal admissions where the principal diagnosis code for the admission was a DFD condition and the rest (64%) reporting admissions where any (principal or additional) diagnosis code for the admission was a DFD condition. Incidence rates were mostly reported as age-adjusted estimates, which were standardised to national or regional populations (64%) or the population studied (21%), with the remainder reported as crude incidence rates (14%).

Table 2 Incidence of hospital admissions for DFD conditions per 1000 person-years with diabetes, unless indicated otherwise
Full size table

Total DFD

Four publications reported admission rates for total DFD conditions combined, in six different populations, with trends over time reported in all (Tables 1 and 2, Fig. 1). Principal admissions were reported in four of these populations (including one in a resident population) and any admissions in two populations. The median (range) initial rate in the five diabetes populations was 16.3 (8.4, 36.6) per 1000 person-years with diabetes and the median (range) end rate was 12.2 (6.4, 20.9). The median (range) relative trend was –24% (–43%, –4%), with trends in four (80%) of the five populations decreasing and the trend stable in one (20%) population. Comparing principal admission rates with any admission rates, the initial rate was similar, the end rate for principal admission rates was lower and the trend for principal admission rates was greater (16.3, 10.5, –24% vs 17.1, 16.1, –7%, respectively).

Fig. 1
figure 1

National trends since 2001 in hospital admission rates for DFD conditions per 1000 person-years with diabetes. When rates are limited to specific ages, these are shown in years in brackets. Total: total admissions for DFD conditions, typically including total aggregated admissions for diabetes-related foot ulcers, infections, PAD and neuropathy. Ulceration: admissions for diabetes-related foot ulceration, typically including ulcers of the foot, ankle or lower leg. Infection: admissions for diabetes-related foot infection, typically including cellulitis and osteomyelitis. PAD: admissions for diabetes-related PAD, typically including PAD diagnosis or peripheral revascularisation procedures. Data are from population-based studies reporting age-standardised incidence rates of hospital admissions for DFD conditions in national populations with diabetes over time. The figure is intended to aid interpretation of relative trends over time between reported initial and end rates; it is not reflective of exact annual trends and should not be used to compare rates between countries, as publications may have used different definitions for diagnosis of diabetes and DFD conditions. If multiple publications from the same country for the same condition were identified, only the publication with the most recent years of data was included; if multiple publications included the same years of data, the publication covering more of the population or condition was favoured (e.g. lower extremity infection was favoured over foot infection). aUnadjusted rate. This figure is available as part of a downloadable slideset

Full size image

Ulceration

Seven publications reported admission rates for ulceration conditions, in eight populations, with trends reported in four (Tables 1 and 2, Fig. 1). Principal admissions were reported in three diabetes populations and any admissions in five (including one in a resident population). The median (range) initial rate in the seven diabetes populations was 5.1 (1.3, 7.6) per 1000 person-years with diabetes and the median (range) end rate was 4.3 (0.6, 7.0). The median (range) relative trend was –7% (–55%, +38%), with trends decreasing in two (50%) and increasing in two (50%) of the populations studied.

Infection

Three publications reported admission rates for infection conditions, in six populations, with trends reported in all (Tables 1 and 2, Fig. 1). All reported principal admissions. The median (range) initial rate was 5.6 (3.8, 9.0) per 1000 person-years with diabetes and the median (range) end rate was 5.1 (2.0, 6.8). The median (range) relative trend was –15% (–57%, +30%), with trends decreasing in three (50%), stable in one (17%) and increasing in two (33%) populations.

PAD

Four publications reported admission rates for PAD conditions, in six populations, with trends reported in three (Tables 1 and 2, Fig. 1). Principal and any admissions were reported in three populations each. The median (range) initial rate was 2.5 (0.9, 3.1) per 1000 person-years with diabetes and the median (range) end rate was 0.7 (0.4, 4.6). The median (range) relative trend was –6% (–76%, +64%), with trends decreasing in two (67%) populations and increasing in one (33%) population.

Amputations

We found 66 publications reporting the rates of admissions for amputation, including 42 reporting total, 39 reporting major and 27 reporting minor amputations (Tables 1 and 3–5, ESM Table 4). Of the 66 publications, most (80%) reported rates using the population with diabetes as the denominator, with the remainder (20%) reporting rates among residents. Nearly all (92%) reported rates for high-income countries, with the rest (8%) reporting rates for middle-income countries. No publications reported rates for low-income countries. Sixteen European, five Asian, three Western Pacific and three North American countries were included, with four publications reporting rates for up to 31 member countries of the Organisation for Economic Co-operation and Development (OECD). Population denominators were identified from national health insurance databases (32%), national diabetes prevalence surveys applied to national census databases (29%), national diabetes registries (20%) or national census databases (20%). Amputation incidence numerators were identified using procedural codes in national hospital discharge databases (80%), national inpatient samples (9%), national health insurance databases (8%) or national surgical theatre records (3%). All publications reported a mutually exclusive incidence of admission for the highest amputation level that occurred in the admission, with 68% including amputation procedures of any cause in people with diabetes and the rest (32%) excluding those caused by trauma or malignancy. Incidence rates were mostly reported as age-adjusted estimates, which were standardised to national or regional populations (48%) or the population studied (18%), with the remainder reported as crude incidence rates (33%).

Table 3 Incidence of hospital admissions for total diabetes-related amputations per 1000 person-years with diabetes, unless otherwise indicated
Full size table

Total amputations

We found 42 publications reporting admission rates for total diabetes-related amputations. In total, 35 reported rates among populations with diabetes and 12 reported rates among resident populations (with four reporting on both) (Table 3). Of the 35 publications reporting rates in 40 different populations with diabetes, 26 reported trends over time in 30 different populations (Tables 1 and 3). The median (range) initial rate was 3.1 (1.4, 10.3) total amputation admissions per 1000 person-years with diabetes and the median (range) end rate was 2.5 (0.6, 7.8). The median (range) relative trend was –27% (–73%, +25%), with trends in 24 (80%) of the 30 populations decreasing, trends in two (7%) stable and trends in four (13%) increasing. In four publications primarily reporting rates stratified by sex, male populations had greater initial and end rates, but the relative trend over time was similar in male and female populations (6.2, 4.1, –50% vs 3.6, 2.0, –51%, respectively) (Table 3).

Of the 12 publications reporting rates in 14 different resident populations (residents with or without diabetes), trends over time were reported for ten populations (Table 3). The median (range) initial rate was 0.155 (0.088, 2.700) total diabetes-related amputation admissions per 1000 residents and the median (range) end rate was 0.135 (0.050, 0.255). The median (range) relative trend was –17% (–44%, +40%), with trends in seven (70%) of the ten populations decreasing and trends in three (30%) increasing. One publication reported rates for 26 OECD member countries, including a mean initial rate of 0.132 total diabetes-related amputation (excluding toe) admissions per 1000 adult residents and a mean end rate of 0.078. Of trends reported for 24 of the countries, the mean relative trend was –41%, with trends in 14 (58%) of the 24 countries decreasing, trends in five (21%) stable and trends in five (21%) increasing [28].

Major amputations

We found 39 publications reporting admission rates for major diabetes-related amputation, including 31 reporting rates among populations with diabetes and 12 reporting rates among resident populations (with four reporting both) (Table 4). Of the 31 publications reporting rates in 36 populations with diabetes, 25 reported trends over time in 30 populations (Tables 1 and 4, Fig. 2a,b). The median (range) initial rate was 1.2 (0.2, 4.2) major amputation admissions per 1000 person-years with diabetes and the median (range) end rate was 0.9 (0.2, 1.8). The median (range) relative trend was –35% (–91%, +10%), with trends in 24 (80%) of the 30 populations decreasing, trends in five (17%) stable and trends in one (3%) increasing.

Table 4 Incidence of hospital admissions for major diabetes-related amputations per 1000 person-years with diabetes, unless otherwise indicated
Full size table
Fig. 2
figure 2

National trends since 2001 in hospital admission rates for major amputations per 1000 person-years with diabetes in (a) all ages and (b) adults by country. Data are from population-based studies reporting age-standardised incidence rates of hospital admissions for major amputation in national populations with diabetes over time. The figure is intended to aid interpretation of relative trends over time between reported initial and end rates; it is not reflective of exact annual trends and should not be used to compare rates between countries, as publications may have used different definitions for diagnosis of diabetes and amputation procedures. If multiple publications from the same country were identified, only the publication with the most recent years of data was included; if multiple publications included the same years of data, the publication covering more of the population was favoured (e.g. type 2 diabetes was favoured over type 1 diabetes). aIncludes only the first amputation; bunadjusted rate. This figure is available as part of a downloadable slideset

Full size image

Of the 12 publications reporting rates in 14 resident populations, trends over time were reported for nine populations (Table 4). The median (range) initial rate was 0.069 (0.005, 1.030) major diabetes-related amputation admissions per 1000 residents and the median (range) end rate was 0.064 (0.002, 0.092). The median (range) relative trend was –26% (–69%, +42%), with trends in seven (78%) populations decreasing, trends in one (11%) increasing and trends in one (11%) stable. One publication reported rates for 31 OECD member countries, including a mean initial rate of 0.074 per 1000 adult residents and a mean end rate of 0.064. Of trends reported for 29 of these countries, the mean relative trend was –14%, with trends in 17 (59%) countries decreasing, trends in seven (24%) stable and trends in five (17%) increasing [21].

Minor amputations

We found 27 publications reporting admission rates for minor diabetes-related amputations, including 25 reporting rates among populations with diabetes and six reporting rates among resident populations (with four reporting both) (Table 5). Of the 25 publications reporting rates in 30 populations with diabetes, 22 reported trends over time in 27 populations (Tables 1 and 5, Fig. 3a,b). The median (range) initial rate was 1.6 (0.3, 4.3) minor amputation admissions per 1000 person-years with diabetes and the median (range) end rate was 1.4 (0.3, 3.8). The median (range) relative trend was –13% (–59%, +49%), with trends in 14 (52%) of the 27 populations decreasing, trends in four (15%) stable and trends in nine (33%) increasing.

Table 5 Incidence of hospital admissions for minor diabetes-related amputations per 1000 person-years with diabetes, unless otherwise indicated
Full size table
Fig. 3
figure 3

National trends since 2001 in hospital admission rates for minor amputations per 1000 person-years with diabetes in (a) all ages and (b) adults by country. Data are from population-based studies reporting age-standardised incidence rates of hospital admissions for minor amputation in national populations with diabetes over time. The figure is intended to aid interpretation of relative trends over time between reported initial and end rates; it is not reflective of exact annual trends and should not be used to compare rates between countries, as publications may have used different definitions for diagnosis of diabetes and amputation procedures. If multiple publications from the same country were identified, only the publication with the most recent years of data was included; if multiple publications included the same years of data, the publication covering more of the population was favoured (e.g. type 2 diabetes was favoured over type 1). aIncludes only the first amputation; bunadjusted rate. This figure is available as part of a downloadable slideset

Full size image

Of the six publications reporting rates in seven resident populations, trends over time were reported for six populations (Table 5). The median (range) initial rate was 0.092 (0.009, 0.123) minor diabetes-related amputation admissions per 1000 residents and the median (range) end rate was 0.098 (0.005, 0.274). The median (range) relative trend was +11% (–49%, +88%), with trends in two (33%) populations decreasing, trends in one (17%) stable and trends in three (50%) increasing.

Discussion

This narrative review analysed findings from 71 eligible publications on hospital admission rates for DFD conditions and amputations in national populations. The findings suggest that national hospital admission rates are substantially higher for DFD than for amputation alone. Trends in admission rates for major amputations are largely decreasing, but trends for minor amputations are inconsistent, decreasing in some countries but increasing in others. Trends in admission rates for DFD conditions are also inconsistent and reported in far fewer publications than for amputations. Although all publications reported the incidence of admissions exclusively for a condition or procedure or set of conditions or procedures, which helps avoid overestimation, there was a high level of heterogeneity in the methods employed to identify admissions and populations. While this review provides the most comprehensive global picture to date on hospital admission rates for DFD conditions and amputations, the conclusions are limited by this heterogeneity and the lack of publications from low- and middle-income countries.

Hospital admissions for DFD conditions

Unlike hospital admission rates for amputations and other diabetes complications [20,21,22,23], hospital admission rates for DFD conditions have only recently begun being used as an outcome for monitoring the burden of DFD [8,9,10, 18]. To our knowledge, this is the first review that has focused on hospital admissions for DFD conditions. The median initial admission rate for a principal or any diagnosis of total DFD conditions was 16.3 per 1000 person-years with diabetes, based on findings from only four publications [32, 35,36,37]. This rate is in keeping with the aggregated median initial admission rates found across ten additional publications on individual DFD conditions, including 5.1 for ulceration, 5.6 for infection and 2.5 for PAD (Tables 1 and 2). However, trend patterns differed, with principal admission rates for total DFD conditions largely decreasing in most populations, but increasing or decreasing for individual DFD conditions, suggesting that admission rates for some individual DFD conditions are decreasing more rapidly than rates for others are increasing.

These indicative findings of high principal admission rates for DFD conditions should, however, be of immediate concern. We can relatively safely assume that these admissions were primarily caused by DFD conditions. The principal admission rates found for DFD conditions in this review are at least equivalent to rates reported for other major diabetes complications [16, 23, 35, 37,38,39]. This supports recent studies suggesting that DFD is a leading cause of diabetes-related hospital admissions in high-income countries [8, 9, 18, 33]. Further, we reviewed hospital admission incidence and not length of hospital stay, which is often used as a measure of condition severity [7]. With DFD also known to result in comparatively long hospital stays, these results are perhaps even more concerning [22, 40, 41].

However, because of the limited number of publications identified, the high level of heterogeneity in hospital diagnosis coding criteria used to identify admissions, including the use of principal or any diagnosis codes and ICD-9 or ICD-10 diagnosis codes, and the different DFD conditions investigated, more research is needed to confirm the high rates and decreasing trends seen for principal admissions for DFD conditions. Ideally, globally recognised standardised and validated ICD coding sets should be developed. In addition, we suggest that admission rates for both DFD conditions and amputations should be reported in future global and national reports of trends in the incidence of diabetes complications [21, 28, 42].

Hospital admissions for amputations

To our knowledge, this is the largest review to date of hospital admission rates for amputation in terms of publications reviewed [15,16,17] and the first in over a decade to primarily investigate global trends in diabetes-related amputations [17]. In this review of mostly 21st century rates, we found a median initial rate of hospital admission for total amputations of 3.1 per 1000 person-years with diabetes (Tables 1 and 3), which is lower than the rate of 5.9 found in the previous global review, which reported mostly 20th century rates [17]. Although most publications included in this review reported total amputation rates, this outcome is no longer recommended; instead, it is recommended that major and minor amputation rates are reported separately, as they are performed for different clinical reasons [14, 16, 42].

For major amputation, we found a median initial hospital admission rate of 1.2 per 1000 person-years with diabetes (Tables 1 and 4), which is lower than the rate of 2.3 found in the previous review [17]. Similar to other reviews [15,16,17], we found decreasing trends in major amputation rates over time among most (80%) populations studied. This may suggest that there have been gradual improvements in DFD care globally. However, given that most publications did not report nationwide improvements in DFD care and the inconsistent trends found in the only publication that reported rates for multiple (OECD) countries [21], these findings may equally reflect changes in general diabetes care, differing methods used to diagnose and define the population with diabetes over time, resulting in more rapid increases in the sizes of populations with diabetes used as denominators [43], or selection bias in countries potentially ‘self-reporting’ favourable trends. Thus, further regular standardised reporting of major amputation rate trends across multiple countries is required, such as that undertaken by OECD countries, to better understand trends over time [21, 28, 42].

For minor amputation, we found a median initial hospital admission rate of 1.6 per 1000 person-years with diabetes (Tables 1 and 5), which was higher than the rate of 1.2 found in the previous review [17]. In contrast to other reviews [15,16,17], we found inconsistent trends in minor amputation rates over time among the populations studied (52% decreasing, 33% increasing, 15% stable). Further, we found that only 14% of populations had trends that aligned with the hypothesis that improved DFD treatment in hospital would result in increasing minor amputation rates and decreasing major amputation rates. In contrast, 52% of populations showed decreasing trends for both major and minor rates and the remainder showed a mix of trends. An alternative measure proposed as an indicator of improved hospital DFD treatment and outcomes is a decreasing high to low (hi–low) amputation ratio (i.e. major to minor amputation ratio) [44]. When the hi–low amputation ratios were calculated from the amputation rate findings in this review, nearly all (93%) populations were found to have decreasing hi–low amputation ratios. Thus, it should be noted that improved DFD care may result in decreasing trends in both major and minor amputation rates, in which case the hi–low amputation ratios may not be as sensitive to improved DFD care as previously thought [14, 44]. Regardless, further investigation of standardised hospital admission rates for minor amputations, preferably alongside those for DFD conditions and major amputations, is required to determine the impact of DFD care on these rates and ratios.

Limitations

This review has several limitations. First, this was a narrative review. Although we performed a systematic search to identify peer-reviewed publications, grey literature was not included and therefore our findings may be subject to selection/publication bias, as grey literature and peer-reviewed publications have reported different hospital admission rates for DFD conditions and amputations in similar national populations [29, 45]. However, a global search of the grey literature was beyond our scope and it is suggested that this is carried out in future research investigating admission rates in smaller regions. Second, although we summarised the overall methodological limitations, we did not perform quality assessments of the included publications, which increases the risk that our findings are subject to bias. While differences in diabetes and DFD care delivered to the populations included are likely to have contributed to some of the trends, potential (rapid) increases in the sizes of populations with diabetes as a result of improved national screening and increases in diabetes prevalence, differences in methods used to identify admission rates for DFD conditions and amputations (including the use of principal vs additional reasons for admission and ICD-9 vs ICD-10 diagnosis codes), differences in the methods used to standardise rates for different population structures over time, and selection bias owing to authors potentially publishing favourable trends may also have contributed to the trends seen [14, 16, 43]. Third, most studies relied on hospital discharge database diagnosis or procedure codes, which have been shown to considerably underestimate admission rates for DFD conditions and to be less accurate for individual DFD conditions than for DFD conditions and amputations combined [10, 41, 46,47,48]. Fourth, we used descriptive statistics to summarise findings from publications that typically used subtly different definitions for admissions, populations and calculating rates and thus we urge caution when comparing rates between publications [16, 43]. Fifth, we reported only descriptive relative trends, regardless of statistical significance and fluctuations in trends over time and the period of time reported for each publication, and thus the magnitudes of trends are not directly comparable between publications and should be interpreted with caution [14, 16, 43]. Indeed, our choice of 5% as a cut-off for increasing or decreasing trends was arbitrary; this cut-off was used only to facilitate a summary of trends across populations with many different definitions. Finally, we did not find any peer-reviewed publications from low-income countries and found only a few from middle-income countries. Our findings are thus unlikely to be representative of these countries.

Implications

First, the hospital admission rates summarised in this review may be useful for forecasting the resources needed to cater for hospital admissions for DFD conditions and amputations. For example, extrapolating the median principal admission rates of 16.3 and 3.1 per 1000 person-years with diabetes for total DFD conditions and total amputations, respectively, to the approximately 536 million adults with diabetes worldwide [2] suggests that there are an estimated 8.7 million hospital admissions for DFD conditions and 1.6 million for amputations annually worldwide, similar to previous estimates [3,4,5, 9]. However, as our findings are based on publications from predominantly high-income countries, and as higher rates were found in the few publications identified from middle-income countries [49,50,51,52,53], these rates are likely to be underestimates.

Second, while the summarised trends in hospital admissions for DFD conditions and amputations should be interpreted with caution, they do provide useful insights [14, 39, 43]. For example, overall trends in admission rates for major amputation seem to be largely decreasing in high-income countries, yet trends in admission rates for minor amputations and DFD conditions are largely inconsistent. However, as mentioned previously, the reasons for these differences in trends are likely to be multiple and include differences in methods used to identify admission rates for DFD conditions and amputations and to identify diabetes and resident populations, differences in the sociodemographic structures of the populations studied, differences in diabetes and DFD care delivered to those populations and selection bias as a result of authors potentially publishing favourable trends [14, 39, 43]. Because of methodological limitations across publications, we were unable to determine the exact impact that these differences had on summarised trends. However, the trends are based on the best evidence available and we provide recommendations in the following section on standardising methodology to enable the reasons for DFD admission trends to be better understood in the future.

Third, this review identified important risk factors for hospital admission for DFD conditions and amputations. Male populations had higher hospital admission rates for DFD conditions and amputations than female populations across all studies in which sex was stratified [23, 37, 54,55,56,57,58,59]. This has also been found for other DFD outcomes, such as neuropathy and ulceration incidence [3, 5, 41]. Further, publications stratifying for age reported that DFD burdens are increasing most rapidly among younger age groups (<40 years) [10, 24, 32, 60]. Additionally, the few publications that stratified for diabetes type found that those with type 1 diabetes generally had slightly higher rates for both DFD conditions and amputations than those with type 2 diabetes [60, 61]. These outcomes support other recent findings showing that age at diabetes onset and diabetes duration may impact DFD outcomes more than other major diabetes complications [62]. As people with diabetes are being diagnosed younger and surviving longer, this suggests that increasing trends in hospital admissions for DFD are likely without intervention [10, 62, 63].

Recommendations

First, like others [14, 39, 42, 43], we recommend the development of global reporting standards for identifying hospital admissions for DFD conditions and amputations and for standardised calculation of incidence rates. We are aware that several countries have proposed using standard ICD coding sets for identifying admissions caused by DFD conditions [12, 48, 64]. Thus, global DFD groups that have developed other reporting standards and definitions [6, 7] should consider developing a consensus on reporting standards for principal admission rates for DFD conditions. In doing so, we recommend that the IDF guide on diabetes epidemiology studies [43], similar work on OECD amputation rate definitions [42], the proposals from countries for standard or modified ICD coding sets to better identify DFD conditions [12, 48, 64] and the common methodologies identified in this review be considered.

Second, we recommend that any global reporting standards developed are tested for validity and reliability in identifying admissions for DFD conditions against gold standard measures, such as prospective clinical bedside audits. Similar studies have been performed previously and have shown that admission rates identified using ICD diagnosis codes considerably underestimate the numbers of people admitted for DFD conditions [10, 41, 46,47,48]. Thus, such studies are important to ensure that hospital discharge dataset coding is as accurate and efficient as possible to be able to monitor the hospital admission burden caused by DFD.

Third, the impact of different factors that may influence or confound admission rates for DFD conditions, such as age, sex, diabetes duration, age at onset, social deprivation, geographic location, and diabetes and DFD care structures, needs further investigation [12,13,14]. Further, investigating risk factors for admission-related outcomes, such as any DFD admission event and length of stay, compared with risk factors for person-related outcomes, such as first DFD admission event per person and in-hospital mortality, should also be considered in future investigations [14, 32, 43]. We are aware of groups that are currently investigating the influence of such risk factors for both admission-related and person-related outcomes for DFD conditions and amputations in various countries and their impending findings should shed new light on issues to consider adjusting for in future global reporting standards [10, 12, 60, 63].

Finally, we recommend that future studies prospectively investigate large population-based cohorts that are representative of national populations with confirmed diabetes or DFD for risk factors for hospital admission for DFD [43]. One such study has found novel risk factors for admission for DFD in large diabetes cohorts, such as younger age [10]. Further, we are aware of studies being conducted that are investigating the risk factors for principal admissions for DFD in large DFD cohorts, as well as the costs of the hospital burden of DFD [18, 63, 65, 66]. Thus, we suggest that these and other hospital admission outcomes, such as length of stay, admission with and without amputation procedures, and mortality, should continue to be investigated to better understand the DFD hospital burden in the future.

Conclusions

This review provides the largest and most comprehensive global picture of trends in hospital admission rates for DFD and amputations. The findings highlight that DFD is a leading cause of hospital admission. The evidence included in this review suggests that the incidence of hospital admissions for DFD conditions is comparable to that for other major diabetes-related complications and is much higher than the incidence of admissions for amputations alone. We also found evidence that major amputation rates have decreased over time, while trends in admissions for DFD conditions and minor amputations were inconsistent. The findings on admission rates for DFD conditions and amputations provide useful estimates that will enable clinicians, researchers and policymakers to benchmark contemporary national incidence rates. Finally, we recommend that global reporting standards for identifying and monitoring principal admission rates for DFD and amputations are developed in order to better understand the seemingly large impact that DFD has on the global burden of disease caused by diabetes and for use as a starting point to develop programmes to reduce this burden.