International comparison of equity gradients in hospitalizations for ambulatory care sensitive conditions

Abstract

Background

Hospitalizations for Ambulatory Care Sensitive Conditions (ACSCs) are thought to be avoidable with effective preventive care and early disease management, usually delivered in community-based ambulatory care settings. These hospitalizations are both costly and sensitive to the socioeconomic situation of patients. Understanding whether different health systems mitigate the sensitivity of ACSCs to socioeconomic gradients may direct attention to equity-improving approaches to ambulatory health care. This study sought to identify differences in the gradient of Ambulatory Care Sensitive Conditions (ACSC) relative to socioeconomic status across nine countries, namely Australia, Canada, England, Finland, France, New Zealand, Spain, Switzerland, and the United States (US).

Methods

The nine countries participating in the International Collaborative on Costs, Outcomes, and Needs in Care (ICCONIC) calculated age—and sex-specific ACSC hospitalization rates between January 1, 2019, and December 31, 2019, based on data from national data sources on hospitalization and sociodemographic information at small-area levels. We used a common definition of ACSC indicators developed through a literature review of indicators routinely used in healthcare performance monitoring. We employed linear regression models to assess inequalities in ACSC association with area-level socioeconomic deprivation and income using the Slope Index of Inequality and the Relative Index of Inequality.

Results

A persistent pattern higher ACSC hospitalization rates for individuals in most disadvantaged areas remained irrespective of the socioeconomic status measure used. There was, however, considerable variation in relative equity gradients in ACSC hospitalizations between countries. Using the nine condition ACSC definition, the greatest inequalities were observed in England (RII = 0.26), and the least inequalities in ACSC hospitalizations were observed in Spain (RII = 0.68) in the 18 + cohort.

Conclusions

A consistent equity gradient in ACSC hospitalizations was observed across high-income countries included in this study. Addressing these inequities with targeted policies can reduce the cost of care while improving access to high quality care.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-026-14406-4.

Introduction

Avoidable hospitalizations, or hospitalizations for Ambulatory Care Sensitive Conditions (ACSC), have been internationally recognized as a measure of quality and adequacy of ambulatory and primary health care [1–10]. Observational studies have repeatedly shown that ACSC hospitalizations are sensitive to accessibility of ambulatory care but also to socioeconomic status (SES) differences. This relationship between SES and ACSC hospitalizations has been observed using individual-level measures [8, 11–13] and area-level measures [1, 7, 14–20]. For example, a Canadian study demonstrated that individual and neighbourhood SES separately and jointly are associated with an increased risk of avoidable hospitalization [10]. There are several potential underlying mechanisms for the observed equity gradient for ACSC hospitalizations [17, 21]. Lower SES individuals may experience unique barriers to access primary care that directly impact the management of chronic conditions and increase the risk of hospitalization [10, 22]. Previous research suggests that patients from lower SES areas may face shortages of primary care physicians, impacting continuity of care [21], and receive fewer referrals for specialist consultations [17, 23, 24]. Additionally, the lack of adequacy and coordination of social services in low SES areas can impact hospitalization rates [17]. Area-level socioeconomic characteristics are appropriate to explore these associations, as aggregate SES characteristics are believed to reflect the nature of the social environment in which people live [23, 25].

The associated system-level costs of ACSC hospitalizations are recognized to be significant [3]. The average national cost of avoidable hospitalizations for chronic conditions in 30 OECD countries was estimated to be $843 million US dollars in 2018 [26] and in 2017, hospitalizations for ACSC in Portugal had a total cost of 352 million euros [27]. Previous studies have identified international variation in care, costs and outcomes amongst individuals with multiple chronic conditions [28–33] however we could not find any international studies examining cross-country analysis in SES-related variation ACSC hospitalizations. Comparisons across health systems may offer new knowledge regarding cost and quality differences and create opportunities for improvement internationally [28]. The aim of this study is to identify differences in socioeconomic gradient of ACSC hospitalizations across nine countries.

Methods

Country datasets

Nine countries which are part of ICCONIC research collaborative participated in this study: Australia, Canada, England, Finland, France, New Zealand, Spain, Switzerland, and the United States (US). The participating countries used a range of datasets, including administrative claims data, survey data, and registry data [28]. The datasets differed regarding the representativeness of the national population, (Table 1) but all were able to measure a comparable relative SES gradient within the observed data. Participating countries extracted ACSC identified at hospital episode level by age, sex and patients’ socioeconomic level assigned using their place of residence at small-area level. Each country identified patients hospitalized with a primary diagnosis of an ACSC between January 1, 2019 and December 31, 2019. Sample population included the entire at-risk population aged 18 or over except for in Australia and the US where the dataset included only individuals aged 65 and above.

Table 1.

Data sources for all 9 countries

Australia
Total Country Population	25.5 M
Sample Population and Method	n = 267,357 (Survey Respondents in New South Wales) Participants aged 65 + drawn from study participants in Sax Institute’s 45 and Up study. Participants sampled from the Services Australia’s Medicare enrolment database in New South Wales (NSW) with near-complete coverage of the NSW population. People aged > 80 years and those in rural areas were oversampled.
Data Source for Hospitalization	The participants are probabilistically linked to the Admitted Patient Data Collection (APDC) in NSW.
Data Source for SES1	Index of Relative Socio-economic Disadvantage (IRSD) of Socio-Economic Indexes for Areas (SEIFA)
Canada
Total Country Population	38.9 M
Sample Population and Method	n = 12.4 M (Entire province of Ontario; 33% of country) Registered persons database including all individuals with coverage from universal public insurance in Ontario, Canada.
Data Source for Hospitalization	Canadian Institute for Health Information Hospital Discharge Abstract Database
Data Source for SES	Marginal Deprivation Index based on Statistics Canada Census
England
Total Country Population	56.3 M
Sample Population and Method	n = 44.3 M (Entire 18 + population) Population estimates in England for adults aged 18 + by Index of Multiple Deprivation from the Office of National Statistics.
Data Source for Hospitalization	Hospital Episode statistics Admitted Patient Care Database (HES APC). This contains all admissions to National Health Service (NHS) hospitals in England (secondary care) that required a hospital bed.
Data Source for SES	Office of National Statistics.
Finland
Total Population	5.5 M
Sample Population and Method	n = 5.5 M The FOLK-longitudinal data modules on personal data from Statistics Finland and the Care Register for Health Care maintained by the Finnish Institute for Health and Welfare, inclusive of whole population.
Data Source for Hospitalization	Care Register for Health Care (CRHC) maintained by the Finnish Institute for Health and Welfare
Data Source for SES	The FOLK-longitudinal data modules on personal data maintained by Statistics Finland
France
Total Population	67.4 M
Sample Population and Method	n = 50.6 M Administrative claims data from the Système National des Données de Santé (National Health Data System) covering the entire population in France (universal health coverage).
Data Source for Hospitalization	Administrative claims data from the National Health Insurance Fund
Data Source for SES	INSEE (National Institute of Statistics and Economic Studies)
New Zealand
Total Population	3.8 M
Sample Population and Method	3.8 M All data for NZ were drawn from the integrated data infrastructure (IDI) a national collection of linked administrative and survey data for full population. Not official government statistics. They have been created for research purposes from the Integrated Data Infrastructure (IDI), which is carefully managed by Stats NZ. For more information about the IDI, please visit https://www.stats.govt.nz/integrated-data/.
Data Source for Hospitalization	National Minimum Data Set (NMDS) – full population administrative data set of publicly funded hospital admissions
Data Source for SES	Total annual income derived from tax data and the NZ Deprivation Index 2018.
Spain
Total Population	47.4 M
Sample Population and Method	n = 41.8 M Atlas VPM data infrastructure, is curated and maintained by the Data Science for Health Services and Policy Research Unit at the Institute for Health Sciences in Aragon, Spain. It represents 88.2% of the Spanish population. These are not official government statistics.
Data Source for Hospitalization	Minimum Basic Data Set at Hospital Discharge (MBDS) for the National Health System in Spain
Data Source for SES	National Statistics Institute (INE) - Living Conditions Survey
Switzerland
Total Population	8.6 M
Sample Population and Method	n = 1.1 M (hospitalized patients only) Administrative, patient-level hospital admission and discharge data provided by the Swiss Federal Statistical Office (FSO) containing 100% of national hospital inpatient admission and discharge data in Switzerland.
Data Source for Hospitalization	Federal Statistical Office (FSO): Statistique médicale des hôpitaux/ Medizinische Statistik der Krankenhäuser
Data Source for SES	Official comprehensive census data obtained from the Federal Statistical Office (FSO), aggregated at the MedStat area level by the Swiss company, MicroGIS SA. [data only available for hospitalized patients without full population -- not used in study]
United States
Total Population	328.2 M
Sample Population and Method	n = 32,590,600 Medicare fee-for-service (FFS) data, 100% sample of all patients age 65 years or older. 63% of the over 65 population covered in FFS program.
Data Source for Hospitalization	CMS Research Identifiable Files - Medicare Claims
Data Source for SES	Calculated from US census variables from the 5-year American Community Survey.

¹Socioeconomic Status (SES)

ACSC hospitalization measures

For the purpose of this study, two alternative definitions of ACSCs were used. The primary definition of ACSCs was developed by reviewing available and commonly used definitions, including country-specific ACSC definitions routinely used in performance monitoring in participating countries. Our resulting primary definition of ACSCs included the following nine health conditions: angina, asthma, congestive heart failure (CHF), convulsions and epilepsy, chronic obstructive pulmonary disease (COPD), diabetes, hypertension, pneumonia, and urinary tract infections. The second definition was provided by the Organisation for Economic Cooperation and Development (OECD) Healthcare Quality and Outcomes indicators list [34] and included five of the nine health conditions identified by our primary definition: asthma, CHF, COPD, hypertension and diabetes. This second definition was employed as a sensitivity analysis. The total number of hospitalizations was provided by all countries and included all hospitalizations for any of the ACSCs as the primary diagnosis (except where exclusions were applied in the secondary diagnoses) to provide an indicator of the total burden of the conditions on the health system. A complete reference of ICD-10 codes for all selected ACSC health conditions can be found in Supplementary Table S1. All countries provided hospital data for all conditions.

Socioeconomic measures

We included two approaches to estimate SES: (1) income quintile (local area-level median household income or similar) and (2) deprivation index quintile (material or social deprivation). This combination of SES indicators was previously used by Wallar & Rosella [10] in examining the risk of index hospitalization by chronic ACSC condition. Including multiple measures enabled a more robust analysis. Based on data availability and comparability across participating countries, the income quintile was selected as the primary measure of SES, while the deprivation index was used in our sensitivity analysis. The material deprivation elements included by each country are listed in Supplementary Table S2. The measurement unit for area-level indicators varied by country depending on national standards with the target population size of 500–1500 individuals. Country-specific area-level measurements and specific measures included in the deprivation indices are reported in Supplementary Table S2.

Demographic variables

Rates were reported by age group (18–39 years, then 5-year groupings up to a 90 + age category) and by sex; Switzerland included those aged 15 + years based on their national age-group standard.

Analysis of inequality

We measured inequalities in ACSC hospitalizations using the slope index of inequality (SII) and the relative index of inequality (RII). The use of these two indices is well described in the health equity literature [35–37]. These two indices were used to quantify the absolute and relative inequality gap, respectively [35, 36]. The SII is interpretable in the scale of the measure of interest [35, 36], in this case, the ACSC hospitalization rate. Used in this way, SII indicates the difference in ACSC hospitalization rates between individuals in the lowest income/deprivation quintile and those in the highest. Higher SII values indicate a greater difference in hospitalization ratess and a greater difference in the absolute equity gap. In contrast, the RII is dimensionless and expresses inequality in relative terms [35, 36]. This allows RII to be comparable across indicators but causes it to be less interpretable for a specific indicator [35, 36]. An RII value < 1 indicates relative inequality between the lowest and highest income/deprivation quintiles, with smaller values (closer to 0) indicating the greatest relative inequality.

The database for the regression consisted of an age-sex-SES specific count of the total number of ACSC hospitalizations and the total at-risk population. Both the SII and the Relative Index of Inequality RII were derived using a linear regression model, which can be expressed as Y = β₀ + β₁X. In the model, Y represents the age-sex adjusted ACSC rate for each SES group, and X represents the cumulative percentage of the population ranked by SES (i.e., the RIDIT score). The SII is given by the coefficient of the linear regression, β₁, while the RII is calculated by dividing the SII by the population mean [37].

All aggregated data were standardized by age and sex, using the Ontario population as the reference. Calculations were conducted at ICES in Ontario (Canada), using SAS Enterprise Guide 8.3 (SAS Institute Inc., Cary, NC, USA). Separate analyses were completed for ages 18 + and 65+, with the latter results including data from Australia and the US. A sensitivity analysis was performed to estimate the inequalities in ACSC hospitalizations using the OECD’s ACSC definition (five diagnostic conditions included) and material deprivation as a measure of SES.

Results

Sample characteristics age 18+

Table 2 provides characteristics of the ACSC sample for seven countries for which data is available for ages 18+. Sex and age standardization using the Ontario population mitigated these differences in the comparison of population ACSC rates and equity gradient analyses. The age-sex adjusted rates of ACSC hospitalizations per 100,000 individuals in the 18 + cohort with nine conditions ranged from 979 in Canada to 1,896 in Finland, with Spain (985), Switzerland (1,227), France (1,287), New Zealand (1,386) and England (1,598) filling the entire range between. Each country had a notable gradient in ACSC hospitalization rates by both income and material deprivation quintile, with lower SES accompanied by higher rates of ACSC hospitalizations.

Table 2.

Sample (demographic) characteristics for 7 Countries (18+)

	Canada	England	Finland	France	New Zealand	Spain	Switzerland*
Sample size (n)	12,168,554	44,263,393	4,459,583	50,563,133	3,766,104	35,394,746	7,311,115
Age
Mean (SD)	48.6 (18.4)	48.9 (18.9)	50.4 (19.1)	51.3 (19.7)	47.4 (18.7)	50.3 (18.0)	56.5 (21.2)
Median	48	48	50	51	46	49	57
Proportion female (%)	51	51	51	53	51	52	51
Number Individuals with ACSC (n)	92,275	587,742	73,394	645,254	50,529	295,728	55,914
ACSC hospitalization count (total)	119,321	748,009	97,038	821,376	69,930	373,045	92,346
Population Age amongst ACSC patients
Mean (SD)	71 (16.9)	69.4 (18.9)	73.3 (15.9)	72.4 (18.1)	65.7 (19.7)	73.6 (16.2)	71.8 (16.4)
Median	74	74	77	76	70	77	77
Proportion female amongst ACSC patients (%)	50	54	52	50	52	47	50
Overall crude ACSC rate per 100k	978.9	1,689.9	2,175.9	1,624.5	1,856.8	1,054	1,263.1
Overall age-sex adjusted ACSC rate per 100k	978.9	1,597.9	1,896.2	1,286.5	1,386	984.6	1,227.2
ACSC Rate per 100k by Income Quintile
Q1 (lowest income) (Rate/100k)	1,453.6	2,354.1	3,343.6	1,984.1	1,946.7	1,145.2	1,470.2
Q2 (Rate/100k)	1,082.3	1,828.2	3,626.2	1,852.9	1,540.5	1,286.1	1,356.1
Q3 (Rate/100k)	898.2	1,605.3	1,904.3	1,594.8	1,288.7	1,084.0	1,303.2
Q4 (Rate/100k)	768.8	1,461.5	1,166.5	1,395.7	1,041.1	899.0	1,215.4
Q5 (highest income) (Rate/100k)	668.1	1,227.9	839.1	1,275.1	921.0	809.4	1,154.6
ACSC Rate per 100k by Material Deprivation
Q1 (least deprived) (Rate/100k)	678.6	1,279.0	n.a.	1,273.2	819.7	n.a.	1,102.0
Q2 (Rate/100k)	803.5	1,464.3	n.a.	1,384.7	1,062.0	n.a.	1,217.4
Q3 (Rate/100k)	917.8	1,578.4	n.a.	1,603.9	1,243.3	n.a.	1,270.5
Q4 (Rate/100k)	1,086.6	1,776.6	n.a.	1,804.4	1,539.6	n.a.	1,253.1
Q5 (most deprived) (Rate/100k)	1,411.1	2,383.5	n.a.	2,085.6	2,014.4	n.a.	1,504.7

* Age 15+; ACSC Ambulatory Care Sensitive Conditions; n.a. data not available

Inequality in ACSC hospitalizations: results age 18+

Figures 1 and 2 provide SII and RII results by income quintile for the seven countries with complete data. Greater SII values (absolute difference in hospitalizations between the highest and lowest income quintiles) are accompanied by lower RII values (further distance below the equality rate of 1.0). The largest SII result in the 18 + cohort is for England, with a total slope of 1,955 representing the incremental hospitalization rate amongst individuals living in the lowest income neighbourhoods. The SII for the remaining countries was 1,754 for Finland, 1,603 for New Zealand, 814 for Canada, 689 for France, 373 for Spain and 344 for Switzerland. The SII results for all countries except Spain showed a statistically significant inequality across SES groups. Across countries, Switzerland and Spain had the lowest SII, followed by \England, Finland and New Zealand with France and Canada falling in between. The confidence intervals are not based on the same population and cannot be used to demonstrate significance. The relative inequality results revealed the lowest (most extreme) RII Ratio was 0.26 for England, with the highest being 0.77 for Switzerland. The relative order of ranking was similar to the SII results with the ordering being Spain (0.68), France (0.58), Canada (0.41), Finland (0.36), and New Zealand (0.27).

Fig. 1.

Slope index of inequality by income quintile for 7 Countries (18+). Legend: SII – Slope Index of Inequality. Higher SII values indicate a greater difference in hospitalizations and a greater difference in the absolute equity gap. CI – Confidence Interval

Fig. 2.

Relative index of inequality by income quintile for 7 Countries (18+). Legend: RII – Relative Index of Inequality. An RII value < 1 indicates relative inequality between the lowest and highest income/deprivation quintiles, with smaller values (closer to 0) indicating the greatest relative inequality. CI – Confidence Interval

The results from the sensitivity analysis using material deprivation as the SES indicator are reported in Supplementary Figures S1 and S2. The SII results maintain the same rank order among the compared countries: England (1,920), New Zealand (1,645), Canada (806), France (630) and Switzerland (503). Finland and Spain were not included in this sensitivity analysis as they did not have a comparable and linkable material deprivation scale. The RII results also displayed a similar rank order: New Zealand (0.26), England (0.27), Canada (0.41), France (0.61), and Switzerland (0.68). Overall, the relative scores and rankings are very similar across the different SES measures and equity indices.

ACSC hospitalizations for 65+

The demographics for the 65 + cohort are included in Table 3. As with the 18 + results, we also see the gradients in ACSC according to income, with essentially the same overall ranking (Figs. 3 and 4). Sensitivity analyses for the 65 + age group using material deprivation and restricting to the five OECD standard ACSC measures are included in Supplementary Figures S3 and S4. Additional sensitivity analyses considering nine conditions also provide very similar results as shown in Supplementary Figures S5 and S6 for the 65 + cohort. No notable deviations from the core findings were observed from the sensitivity analysis.

Table 3.

Sample (Demographic) characteristics for 9 Countries (65+)

	Australia	Canada	England	Finland	France	New Zealand	Spain	Switzerland	US
Sample size (n)	42,737	2,587,593	10,353,716	1,204,718	14,138,612	782,835	7,897,314	1,605,800	32,590,600
Age
Mean (SD)	76.9 (7.9)	74.9 (7.8)	75.0 (7.2)	74.6 (7.4)	76.2 (8.4)	74.7 (7.6)	76.0 (7.9)	77.4 (7.5)	74.5(7.3)
Median	76	73	74	73	74	73	74	77	72
Proportion female (%)	47	54	54	56	56	53	57	53	54
Number Individuals with ACSC (n)	4093	64,759	392,271	57,088	461,245	30,345	201,193	39,595	946,627
ACSC hospitalization count (total)	6010	83,189	507,430	76,855	599,284	42,747	283,490	68,590	1,223,016
Population Age amongst ACSC patients
Mean (SD)	80.7 (8.4)	79.94 (8.6)	80.3 (8.3)	80.1 (8.2)	81.8 (9.2)	79 (8.3)	81.0 (8.2)	79.8 (7.4)	79.3 (8.3)
Median	80	80	80	80	83	79	82	82	72
Proportion female amongst ACSC patients (%)	45	52	53	53	51	50	48	51	71
Overall crude ACSC rate per 100k	14,062.8	3,209.6	4,901	6,379.5	4,238.6	5,460.5	3,589.7	4,271.4	3,752.7
Overall age-sex adjusted ACSC rate per 100k	12,705.7	3,209.6	4,828.4	6,649.3	3,810.3	3,921.7	3,410.6	4,176.3	2,503.8
ACSC Rate per 100k by Income Quintile
Q1 (lowest income) (Rate/100k)	20,807.8	4,372.9	7681.3	8923.3	4988.8	5546.6	3792.9	4928.5	3483.2
Q2 (Rate/100k)	22,654.0	3,414.3	5,982.4	7,654.9	4,549.4	4,436.7	4,191.0	4,783.9	3,018.5
Q3 (Rate/100k)	14,271.4	3,013.5	4,733.3	5,326.7	4,120.4	3,754.7	3,728.8	4,643.4	2,668.1
Q4 (Rate/100k)	10,261.8	2,754.6	4,100.5	4,092.1	3,785.4	3,193.4	3,254.9	4,325.8	2,352.6
Q5 (highest income) (Rate/100k)	6,983.3	2,327.5	3,308.2	3,038.0	3,677.7	2,926.1	2,996.7	3,950.8	1,902.3
ACSC Rate per 100k by Material Deprivation
Q1 (least deprived) (Rate/100k)	10,464.4	2,397.5	3,467.7	n.a.	3,703.8	2,499.7	n.a.	3,734.8	1,874.5
Q2 (Rate/100k)	12,629.9	2,864.6	4,105.8	n.a.	3,813.1	3,216.7	n.a.	4,090.3	2,245.5
Q3 (Rate/100k)	13,850.2	3,061.1	4,649.2	n.a.	4,100.3	3,738.5	n.a.	4,392.2	2,635.3
Q4 (Rate/100k)	15,555.8	3,398.5	5,782.7	n.a.	4,423.9	4,531.7	n.a.	4,373.9	3,019.0
Q5 (most deprived) (Rate/100k)	17,819.6	4,319.4	7,778.4	n.a.	5,060.5	5,664.8	n.a.	5,500.6	3,607.6

ACSC Ambulatory Care Sensitive Conditions; n.a. data not available

Fig. 3.

Slope index of inequality by income quintile for 9 Countries (65+). Legend: SII – Slope Index of Inequality. Higher SII values indicate a greater difference in hospitalizations and a greater difference in the absolute equity gap. CI – Confidence Interval

Fig. 4.

Relative index of inequality by income quintile for 9 Countries (65+). Legend: RII – Relative Index of Inequality. An RII value < 1 indicates relative inequality between the lowest and highest income/deprivation quintiles, with smaller values (closer to 0) indicating the greatest relative inequality. CI – Confidence Interval

Condition-specific results

We undertook one additional supplementary analysis to determine whether there were variations in the specific conditions for which individuals were hospitalized across the different countries. We report the number of hospitalizations for each of the nine conditions in the 18 + cohort as Supplementary Table S3 and for 65 + in Supplementary Table S4. Both the counts and the relative percentages of the conditions were fairly consistent across countries, with some notable variations. Overall, CHF and Pneumonia were the most common conditions, followed by COPD and UTI. Angina, Hypertension, Asthma and Convulsions were generally the lowest, with Diabetes falling in the middle. There were a few exceptions. Diabetes-related hospitalizations were particularly low in England and Finland, COPD was particularly high in France and Switzerland; CHF admissions were also relatively low in England, while Angina-related hospitalizations were higher in Australia, New Zealand and France.

Discussion

This international comparison revealed notable differences in ACSC hospitalization rates across countries as well as a consistent and distinct pattern of socioeconomic inequities in ACSC hospitalizations.

This study identified differences in ACSC hospitalization rates across countries but of greater interest are the differences in equity gradients. Higher rates of ACSC hospitalizations were observed in lower SES groups across most participating countries. Likewise, the lowest rates of ACSC hospitalizations were observed in local areas with the highest income quintile across all countries. The observed equity gradient persisted regardless of SES measure (material deprivation or income quintile) or ACSC diagnostic grouping (five or nine conditions). In all configurations, most countries displayed monotonically increasing hospitalizations with higher material deprivation or lower income in the area. In the 18 + population, the SII absolute measure of inequality identified the greatest inequalities in ACSC hospitalizations in England, followed by Finland, New Zealand, Canada, France, Spain, and Switzerland respectively. The relative index of inequality identified a similar relative order rank, except for Finland and New Zealand switching positions. This ordering was similar in the 65 + population except that Australia had a much larger SII. Australia had a very large SII because the overall rate of ACSC in Australia is much higher than other countries. This was also shown in Ansari et al. 2012 [16]. Using their rates and our standardized population we calculated the equivalent overall age-standardized rate amongst those aged 65 + as 12,817.1/100,000 which is very close to the SAX sample estimate of 12,705.7. Australia also had a remarkably high ACSC rates by in the lowest two income quintiles. Both Australia and Spain uniquely had slight aberrations to linearity with highest ACSC rates in the second lowest income quintile, which also contributed to larger confidence intervals for these two countries.

There may be several reasons for the large inequality gap in England. Firstly, the gap may be driven by the prevalence of ACSC health conditions. For instance, in England, COPD accounted for 16.8% of all ACSC admissions for 18 + and 18.9% for 65+. Pneumonia, which may occur more frequently amongst individuals with COPD, accounted for 29.8% in the 18 + cohort in England and 34.2% in the 65 + cohort (the highest proportion amongst all countries) of ACSC hospitalizations in England. This may coincide with the reported upward trend in COPD admissions in England and underdiagnosing in primary care [38], as primary care provided in the community has been identified as a predictor of both COPD-related and all-cause readmission rates [39]. According to both inequality measures, Spain reported the lowest inequalities in ACSC hospitalizations. Spain also has the fourth lowest overall rate of potentially avoidable hospitalizations among EU countries that collect this data and the lowest rate among countries included in this study [40]. The distribution of resources and institutional design of the Spanish National Health System according to the administrative healthcare areas, with all residents registered to a Family Doctor, may contribute to these relatively low rates of avoidable admissions for chronic diseases [41].

The conditions leading to the results demonstrated by our analyses are essential to driving policy action. Factors such as the availability and access to community-based ambulatory care are key components that have been identified to be related to ACSC hospitalizations [1, 11, 20]. Interestingly, primary care spending as a share of total health expenditure by the OECD revealed that the UK and Finland report the highest primary care spending (15.7 and 15.2% respectively) despite their relatively high inequality gradients [40]. In contrast, Canada reported some of the lowest primary care spending amongst participating countries (10.9%), yet displayed relatively lower inequality gradients. We recommend additional research with robust measurement of these exposures to assess associations between gradients and a broad array of country-specific factors such as the availability of inpatient beds or the condition-specific financial incentives to refer or not to secondary care as causal factors related to the differences demonstrated in this study.

A recent OECD health systems report [42] investigated the relationship between primary care system characteristics and avoidable hospital admissions across OECD countries, focusing on admissions for asthma, chronic obstructive pulmonary disease, and congestive heart failure. Interestingly, considering the 65 + population (with all 9 countries represented), the three countries with the highest rates of avoidable hospitalizations - Australia, Finland, and England [43], - demonstrated the greatest inequalities in ACSC hospitalizations in our study. Alternatively, the three countries from the present study that displayed the least inequalities – Spain, France and Switzerland –displayed the second lowest avoidable hospitalizations. Three actionable policy levers defined the clusters: gatekeeping; continuity of care; financial incentives for primary care physicians to improve quality of care. The term gatekeeping refers to the requirement that primary care practitioners serve as the ‘gatekeepers’ to other levels of care, while continuity of care was described as the share of the population that has a regular doctor to go to for care. While both the top performing cluster (including England and Australia) and second cluster (including Spain, France and Canada) were characterized by strong incentives to primary care physicians, there was variability in the level of gatekeeping and continuity of care both between and within clusters. The top performing cluster (England and France) were categorized by strong gatekeeping and strong continuity of care. The second cluster (including Spain, France and Canada were characterized by either limited gatekeeping and strong care continuity or strong gatekeeping and limited care continuity. In reviewing the three countries included in the present analysis, Spain and Canada were identified as having strong gatekeeping with limited care continuity, while the health system in France was identified as having limited gatekeeping and strong continuity of care. Given the inequality results of the present study, the two primary care policy levers of gatekeeping and continuity of care, both individually and in conjunction, may influence the equity-based disparities attached to avoidable hospitalizations [42]. Explicit local comparisons of primary care supply and access, as well as other ambulatory care services, would be a good complement to assess the associations with the differences observed in this study. The present international comparison allows us to identify existing inequities but prompts further study to seek connections with health system organization and equity of care delivery [16, 19]. This study did not assess whether such difference might be most attributed to underlying prevalence, risk factors for exacerbations (e.g. smoking, diet, seasonal infection or immunization rates for influenza, which could exacerbate pneumonia, etc.). Future research should investigate hospitalizations by specific ACSC health condition and country to assess whether certain conditions are more sensitive to socioeconomic gradients.

Limitations

There is variability in the population size and representation amongst the represented countries. National population datasets ranged in size across the nine participating countries. Finland had the smallest population (n = 4,459,583), while France had the largest (n = 50,563,133) in the 18 + cohort. The population median age in the 18 + cohort ranged from 46 in New Zealand to 57 in Switzerland and was very close to the population mean in all countries. The proportion of females with an ACSC ranged from 47% to 54% and varied slightly from the population representation, most notably in Switzerland, where females were less commonly represented in the ACSC hospitalized population. The application of sex and age adjustments addresses difference in age and sex representation across countries.

The variability in ACSC definitions and codes in past studies invites challenges with comparisons across geographies and datasets [44]. This study provides an example of developing a common definition and set of codes that are comparable across a number of countries and regions. By focusing on a hospital-based metric, this study increased the standard measurement of diagnoses across countries. All countries had consistent access to analogous hospitalization data to ensure equivalency in the measurement of ACSC conditions. We present a clear ACSC definition with both OECD 5-condition and a more comprehensive 9-condition measure.

The measurement of SES was more variable across countries. While all countries were able to generate income quintiles, these may be less relevant in some populations, such as youth and older adults who are not fully engaged as wage earners in the economy. Furthermore, other social measures such as crime, sole-parent, unemployment, and education are also determinants of health and associated with access to health services. These are generally included in material or social deprivation measures with some variation across countries. Amongst the countries that did provide a measure of deprivation, there is considerable overlap in the constituent factors. Moreover, the relative gradient is likely more important than the specific constituent items for the purpose of measuring SII and RII in this study. By focusing on the quintiles and the relative inequality gradients as opposed to absolute levels of income or material deprivation, we reduce the impact of definitional variation.

Another limitation is that while the results from each country are representative of the samples from which they are drawn (e.g. regional, insurer, and survey in the case of Australia), they are not all representative of the entire country. The results, however, are representative of the samples from which they are drawn, and these are largely representative of important populations within each country.

Finally, it is important to note that inequalities in ACSC hospitalizations may simply reflect socioeconomic gradients in health status and not in health care access, such that those with the poorest health status have the highest hospital use and expenditure rates [9, 45]. Higher rates of visits and hospitalizations may arise because of higher disease prevalence, increased disease severity, and multiple comorbidities for people experiencing disadvantage [9, 13, 15, 45–47]. There exists an opportunity to reduce ACSC rates if the areas with the lowest SES quintiles could be adequately targeted. However, moving beyond identifying socioeconomic inequities in ACSC hospitalizations to addressing the underlying barriers in low socioeconomic groups remains a complex and multifaceted issue. Previous research has identified common features of successful interventions to reduce ACSC hospitalizations including primary care physician involvement, a high degree of interconnectedness among providers, the provision of care management, and the support of self-management goals [48].

Conclusions

This study confirms a persistent socioeconomic gradient for ACSC hospitalizations common across high-income countries irrespective of socioeconomic measurement approach. We operationalized a literature-informed definition of ACSCs that proved suitable for international comparison. Researchers and policy makers should investigate the underlying drivers of disparities in ACSC hospitalizations and develop targeted policies to improve access to high quality care and reduce overall healthcare costs.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors contributed to the study conception and design. All authors contributed to the analysis for their country’s dataset. YQB conducted the analysis of inequalities. GMS and WPW provided initial interpretations of the analysis of inequalities. The first draft of the manuscript was written by GMS and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This work was partially supported by a Health Foundation (UK) grant. The analyses were supported by the Health System Performance Network, which receives grant funding from the Ontario Ministry of Health and the Ontario SPOR Support Unit and was completed at ICES. Ontario data were compiled and provided by the Canadian Institute for Health Information. The conclusions and opinions expressed herein are solely those of the authors and do not reflect those of the funding or data sources.

Data availability

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

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. The analysis used secondary data sources. Informed consent to participate was obtained from participants at the time of original data collection where required. For the present secondary analyses, the requirement for additional informed consent was waived or deemed not applicable by the relevant institutional or regulatory authorities. In Australia, the 45 and Up Study has ethical approval from the University of New South Wales (UNSW) Human Research Ethics Committee. This study has received ethics approval from the UTS Human Research Ethics Committee (UTS HREC REF NO. ETH18-2507) and the NSW Population & Health Services Research Ethics Committee under reference number 2013/11/487. In Canada, participation in this project was authorized under Sect.  45 of Ontario’s Personal Health Information Protection Act (PHIPA) and does not require review by a Research Ethics Board. In Finland, ethical approval was from the ethical review board of the Finnish Institute for Health and Welfare. In Spain, participation in this project was approved by Ethics Committee for Clinical Research in Aragon (2018-02-28, ref PI17/0411, Acta N 04/2018). Approval to access IDI data for New Zealand for this project was provided by Stats NZ (Ref: MAA2018-35). In the United States, the Institutional Review Boards at both the Harvard T.H. Chan School of Public Health and Brown University approved this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.