If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Deprivation Index and Dependency Ratio predict hospital admission rate.
•
Acute Illness Severity and Chronic Disabling Disease do not predict admissions.
•
Deprivation Index and Dependency Ratio synergistically increase admission rate.
Abstract
Background
Patients from deprived backgrounds have a higher in-patient mortality following an emergency medical admission; there has been debate as to the extent to which deprivation and population structure influences hospital admission rate.
Methods
All emergency medical admissions to an Irish hospital over a 12-year period (2002–2013) categorized by quintile of Deprivation Index and Dependency Ratio (proportion of population <15 or ≥65 years) from small area population statistics (SAPS), were evaluated against hospital admission rates. Univariate and multivariable risk estimates (Odds Ratios (OR) or Incidence Rate Ratios (IRR)) were calculated, using logistic or zero truncated Poisson regression as appropriate.
Results
66,861 admissions in 36,214 patients occured during the study period. The Deprivation Index quintile independently predicted the admission rate/1000 population, Q1 9.4 (95%CI 9.2 to 9.7), Q2 16.8 (95%CI 16.6 to 17.0), Q3 33.8 (95%CI 33.5 to 34.1), Q4 29.6 (95%CI 29.3 to 29.8) and Q5 45.4 (95%CI 44.5 to 46.2). Similarly the population Dependency Ratio was an independent predictor of the admission rate with adjusted predicted rates of Q1 20.8 (95%CI 20.5 to 21.1), Q2 19.2 (95%CI 19.0 to 19.4), Q3 27.6 (95%CI 27.3 to 27.9), Q4 43.9 (95%CI 43.5 to 44.4) and Q5 34.4 (95%CI 34.1 to 34.7). A high concurrent Deprivation Index and Dependency Ratio were associated with very high admission rates.
Conclusion
Deprivation Index and population Dependency Ratio are key determinants of the rate of emergency medical admissions.
Deprivation has been defined by Townsend as a state of “observable and demonstrable disadvantage relative to the local community to which an individual belongs” [
]. There are a number of objective ways of calculating deprivation, in this paper we have utilized the Irish National Deprivation Index for Health and Health Services Research derived by the Small Area Health Research Unit (SAHRU) at Trinity College Dublin [
]. This index utilizes a weighted combination of four indicators, relating to unemployment, social class, type of housing tenure and car ownership to calculate deprivation. The index is reported at the level of the District Electoral Division (DED), the smallest administrative area for which census data are collected. The Republic of Ireland has only recently introduced postal codes, therefore the DED fulfills the same function as a postal code would in other jurisdictions to define small local populations.
Whether ‘the poor’ or those from a deprived background incur a higher demand for acute hospital resources been debated; Epstein et al. suggested that such patients have a longer hospital length of stay and episode costing compared with patients with a less deprived background [
]. A systematic review of the literature found that patients from deprived areas used more emergency care, attended A&E more frequently, appeared to have attended A&E for less serious conditions, accessed outpatient care more via emergency channels and failed to attend a larger proportion of their outpatient appointments [
Are hospital services used differently in deprived areas? Evidence to identify commissioning challenges Oxford: Centre for Health Service Economics & Organisation.
The rate at which patients present with a medical condition requiring an emergency hospital admission is of great interest; hospital admission rates for Ambulatory Care Sensitive Conditions in England in the years 2009/10 varied from 10.1 to 24.5 per thousand and were linearly related to the patient's Deprivation Index [
]. We have expressed the view that emergency medical admissions are substantially determined by Acute Illness Severity Score and Chronic Disabling Disease Score [
Variations and inter-relationship in outcome from emergency admissions in England: a retrospective analysis of Hospital Episode Statistics from 2005–2010.
]; it is tempting to suggest that either institutional infrastructure or the processes of care influence or may explain such difference in outcomes. However, the extent to which the age structure and social deprivation construct of the catchment area are relevant to both admission rates and outcome differences should be investigated [
]. We have related the Deprivation Index and Dependency Ratio (proportion of non-working population < 15 or ≥65 yr) for all emergency medical admissions admitted between 2002 and 2013 to St James' Hospital (SJH) and linked these to the Irish National Census data (Small Area Population Area—the unit being the District Electoral Division level (DED)) to permit ecological relationships to be examined. We have a particular interest as to how emergency admission rates might be correlated with the DED small area population statistics [
SJH serves as a secondary care centre for emergency admissions for its local Dublin catchment area of 270,000 adults. All emergency medical admissions are referred to one of nine teams operating a 24 hour on-call roster. Emergency medical patients are admitted from the Emergency Department (ED) to an Acute Medical Admission Unit (AMAU) opened in 2003, under the care of the on-call physician (who is certified in General Internal Medicine and a Subspecialty)—the operation and outcome of which have been described elsewhere [
]. By design, it is intended that all patients be admitted to the AMAU where their medical needs are assessed. The AMAU is a 59-bedded unit and is staffed by a cohort of experienced nurses who's expertise and sole professional role is confined to the care of acutely ill medical patients. The AMAU is on call 24 h a day and admits approximately 20 patients a day.
2.2 Data collection
For audit purposes we employed an anonymous patient database assembling core information about each clinical episode from elements contained on the Patient Administration System (PAS), the National Hospital In-patient Enquiry (HIPE) scheme, the patient electronic record, the emergency room and laboratory systems. HIPE is a national database of coded discharge summaries from acute public hospitals in Ireland [
]. Ireland used the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) for both diagnosis and procedure coding from 1990 to 2005 and ICD-10-CM since then.
Data held on the database includes the unique hospital number, admitting consultant, date of birth, gender, area of residence, principal and up to nine additional secondary diagnoses, principal and up to nine additional secondary procedures, and admission and discharge dates. Additional information cross-linked and automatically uploaded to the database includes physiological, haematological and biochemical parameters. For this study data was related to all emergency general medical patients admitted to SJH on the database for the years 2002–2013.
Approximately 9.9% of our patients stay >30 days with a median LOS of 54.8 days (IQR 38.8, 97.2)—the majority of these are awaiting nursing home beds or other social interventions rather than ongoing admissions for medical reasons. We have therefore chosen a truncated end-point (death or admission completed by the 30-day endpoint) for analysis, to avoid the additional confounding of non-medical reasons for continuing as an inpatient. Patients temporarily visiting the Republic of Irleand, with permanent addresses outside the jurisdication, were excluded from this analysis. Analysis of admission rate data was confined to the 68% of admissions who resided in our catchment area as those residing elsewhere where more likely to be admitted to other institutions in an emergency.
2.3 Deprivation indices
To classify deprivation, we used the public domain Irish National Deprivation Index for Health and Health Services Research, derived by the Small Area Health Research Unit (SAHRU) at Trinity College Dublin [
]. The SAHRU group employed the census data from the Central Statistic Office (CSO) (1991, updated in 2002, 2006 and 2011) to compute a functional index of deprivation using the District Electoral Division (DED) as the base unit. The Republic of Ireland has 3440 DED; these are the smallest administrative areas for which population statistics are gathered by the CSO. For reasons of confidentiality, sparsely populated areas may be merged; this results in Small Area Populations Statistics (SAPS) being available for 3409 DED. Using principle component analysis (PCA), a weighted combination of four indicators, relating to unemployment, social class, type of housing tenure and car ownership was derived, as described by the SAHRU investigators [
]; the Deprivation Index scores were ranked from low (least deprived) to high (most deprived) and divided into quintiles according to their ranked raw scores. We utilized the registered address on our PAS, to allocate each address to a divisional area, with a corresponding matched SAHRU Deprivation raw score and quintile rank. This attribute data were joined to the small area polygon geometries based upon their relative geographic positions, using the ArcGS Geographic Information System software implementation of the Point-in-Polygon algorithm as outlined by Shimrat [
From the 2006 Census returns information, the SAPS for each small area population unit (the unit being a DED with a median population size of 2845 (IQR 2020, 3399)) was available on the total population within each unit area and the population Dependency Ratio (proportion of non-working population (<15 and ≥65 yr)). The annual incidence rate for emergency medical admissions (expressed/1000 of population) was calculated by summing total admissions by each ED over 12 years and calculating an average for each (numerator) and dividing by the total population within each area (divisor).
2.4 Statistical methods
Descriptive statistics were calculated for background demographic data, including means/standard deviations (SD), medians/inter-quartile ranges (IQR), or percentages. Comparisons between categorical variables and mortality were made using chi-square tests.
For hospital Admission Rates, we employed a truncated Poisson regression model, including predictive outcome categorical variables (e.g. disabling score groups) in the model as a series of indicator variables. The dependent variable of the Admission Rate is a positive integer; it cannot have zero values. The predictor variables are therefore regressed against Admission Rates using the truncated Poisson model. We used robust standard errors for the parameter estimates, as recommended by Cameron and Trivedi [
]. The Poisson regression coefficients are the log of the rate ratio: the rate at which events occur is called the incidence rate. Thus with the truncated Poisson regression model, we can interpret the coefficients in terms of incidence rate ratios (IRR). We used the margins command in Stata 13.1 to estimate and interpret adjusted predictions for sub-groups, while controlling for other variables such as illness severity, using computations of average marginal effects. Margins are statistics calculated from predictions of a previously fitted model at fixed values of some covariates and averaging or otherwise over the remaining covariates. In the multivariable model (logistic or Poisson), we adjusted univariate estimates of effect, using known outcome predictor variables such as the Acute Illness Severity Score [
] and sepsis status (as defined by 3 groups of a) no blood culture requested, b) blood culture requested with negative result and, c) blood culture request with positive result [
]. The model parameters were stored; post-estimation intra-model and cross-model hypotheses could thereby be tested.
Adjusted odds ratios (OR) and 95% confidence intervals (CI) or incidence rate ratios (IRR) were calculated for those predictors that significantly entered the model. Statistical significance at P < 0.05 was assumed. Significance was adjusted for multiple comparisons using Scheffe's method. Stata v.13.1 (Stata Corporation, College Station, Texas) statistical software was used for analysis.
3. Results
3.1 Patient demographics
A total of 66,203 episodes were recorded in 35,631 patients admitted as medical emergencies over the 12 year study period. The proportion of males and females was 48.9% and 51.1% respectively. The median (IQR) length of stay (LOS) was 5.1 (2.1 to 9.8) days. The median (IQR) age was 62.3 (42.1 to 77.2) years, with the upper 10% boundary at 84.5 years.
The frequency of patients with Chronic Disabling Disease Scores of 0, 1, 2, 3 or 4+ was 11.2%, 24.6%, 29.2%, 21.1% and 13.9%; their respective 30-day mortality rates were 1.2%, 4.0%, 7.8%, 13.5% and 25.5%. Charlson Co-Morbidity Index of 0, 1 and 2 occurred in 45.5%, 27.3% and 27.2% of patients; their 30-day mortality rates were 3.0%, 8.9% and 22.2%. Acute Illness Severity Groups I–VI occurred in 2.9%, 7.0%, 11.7%, 15.6%, 18.6% and 35.5% (data not available for 8.8%); their respective 30-day mortality rates were 0.13%, 0.12%, 0.65%, 1.45%, 4.8% and 24.0%. The frequency of the three sepsis groups of a) no blood culture request, b) culture request with negative result and, c) culture request with positive result, was 75.4%, 21.2% and 3.5% of patients, with respective 30-day mortality rates of 5.6%, 16.5% and 32.2%.
3.2 Deprivation Index and admission characteristics (Table 1)
The demographic characteristics of the admission profile by Deprivation Index quintile are shown in Table 1.
The risk of an in-hospital death by day 30 increased with Deprivation Index quintile; the respective mortality rates were 8.4%, 8.2%, 9.7%, 9.2% and 9.2% (p = 0.001). The fully adjusted multivariable ORs, of a 30-day in-hospital death, for Deprivation Index quintiles II–V (vs. quintile I) were 1.12 (95% CI 0.98 to 1.27), 1.35 (95% CI 1.19 to 1.54), 1.34 (95% CI 1.18 to 1.54) and 1.59 (95% CI 1.38 to 1.82) (p < 0.001 for quintiles III–V).
3.3 Admission rates related to Deprivation Index and the Dependency Ratio of the population
The admission rates showed a modest increase with age (Fig. 1). There was little statistical relationship of the admission rate to Acute Illness Severity Score (p = 0.67), Charlson Co-morbidity Index (p = 0.23), Chronic Disabling Disease Score (p = 0.12), or sepsis status (p = 0.81).
Fig. 1Relationship between age group and hospital admission rate for emergency medical conditions. From each electoral division, emergencies were summated over 12 years, then averaged and expressed as a ratio of the total population in the 2006 census. Data was modelled by Poisson regression of hospital admission rates against known outcome predictors including Acute Illness Severity Score (AISS), Chronic Disabling Disease Score, Charlson Co-morbidity Index, sepsis status and Deprivation Index. The marginal effect of age group was then calculated from predictions of the fitted model at fixed values of the abscissa covariate and averaging over the remaining covariates.
Increasing Dependency Ratio and Deprivation Index quintiles increased hospital admission rates as shown in Fig. 2, Fig. 3 respectively. The corresponding Incidence Rate Ratios are shown in Table 2.
Fig. 2Relationship between Dependency Ratio and hospital admission rate for emergency medical conditions, using methodology as in Fig. 1. The marginal effect of Dependency Ratio was then calculated from predictions of the fitted model at fixed values of the abscissa covariate and averaging over the remaining covariates.
Fig. 3Relationship between Deprivation Index and hospital admission rate for emergency medical conditions, using methodology as in Fig. 1. The marginal effect of Deprivation Index was then calculated from predictions of the fitted model at fixed values of the abscissa covariate and averaging over the remaining covariates.
Table 2Truncated Poisson regression model of Hospital Incidence Admission Rates.
Robust
Predictor
IRR
Std. err.
Z
P > z
[95% conf.
interval]
Dependency ratio
Q2
0.93
.01
−7.9
0.000
0.92
0.95
Q3
1.28
.01
28.8
0.000
1.26
1.30
Q4
1.78
.11
76.2
0.000
1.75
1.80
Q5
1.68
.01
76.0
0.000
1.66
1.70
Deprivation quintile
Q2
1.68
.02
40.7
0.000
1.64
1,72
Q3
3.04
.04
87.6
0.000
2.97
3.12
Q4
2.68
.03
80.6
0.000
2.62
2.75
Q5
3.64
.05
106.6
0.000
3.75
3.94
IRR—the incidence rate ratios for the Poisson model—these are derived by exponentiating the Poisson regression coefficient. There are two interpretations of the calculations 1) the log of the ratio of expected counts explaining the “ratio” and 2) counts as number of events per time (or space), hence “rate” in incidence rate ratio.
There was a strong interaction between Deprivation Index and Dependency Ratio; it was apparent that highly deprived areas with a concurrent high dependency ratio were particularly likely to sustain very high hospital emergency medical admission rates (Fig. 4).
Fig. 4Interaction between Dependency Ratio and Deprivation Index using methodology as in Fig. 1. The marginal effect of Dependency Ratio is illustrated for patients from the highest quintile of Deprivation Index—admission rates were particularly high in DED's with a concurrent high Dependency Ratio and Deprivation Index.
The major determinants of hospital admission rates in our cohort were Deprivation Index and the population Dependency Ratio. The admission incidence rates increased with increasing deprivation and with an increasing proportion of non-working population within a given area. Furthermore there was a strong interaction between these two factors which further increased the admission rate. Surprisingly known outcome predictors such as Acute Illness Severity Score, Chronic Disabling Disease Score and sepsis status appeared to play little role in determining hospital admission rates. There was a small effect of increasing Charlson Co-morbidity Index but this appeared of significantly less importance than Deprivation Index and Dependency Ratio.
We, and others, have previously shown that deprivation is an independent predictor of 30-day in-hospital mortality [
]. A study by the Centre for Health Service Economics & Organisation showed that patients from deprived areas utilized emergency care services more frequently and for less severe conditions but did not demonstrate that this was due to deprivation itself [
Are hospital services used differently in deprived areas? Evidence to identify commissioning challenges Oxford: Centre for Health Service Economics & Organisation.
]. The King's Fund reported that admission rates for ambulatory care sensitive conditions were linearly dependent on the underlying deprivation quintile [
]. Previous studies have reported the association of deprivation with admission rates for particular conditions including diabetes, coronary heart disease and chronic obstructive pulmonary disease [
Association of population and primary healthcare factors with hospital admission rates for chronic obstructive pulmonary disease in England: national cross-sectional study.
]. A number of studies have previously reported that deprivation assessed at general practice level was one of a number of predictors of hospital admission rates [
]. This work extends our knowledge of the influence of deprivation by demonstrating that it is a powerful independent predictor of hospital admission rates at a small area level. Indeed in our analysis it was the only modifiable factor which influenced both mortality and admission rate, making it ideal for intervention.
Like any study ours has limitations. The current work evaluated admissions to a single centre with a relatively deprived elderly population. The results require validation in other cohorts to establish their generalizability to the population at large. For practical reasons related to data collection our study excluded patients admitted under services other than general medicine, therefore the results cannot immediately be extrapolated to conditions admitted under these services. A further potential limitation is the relevance of a hospital admission compared to other important outcomes such as mortality. However we would argue that the significant economic costs of avoidable hospital admissions justify intervention to address this as an additional important target for health care improvement.
Our results challenge fundamental concepts in the practice of medicine. Clearly age is not a modifiable factor, deprivation is however. These data suggest that interventions aimed at ameliorating deprivation will have a profound influence on the rate of hospital admissions, in particular in elderly populations. This approach has a parallel in public health approaches such as vaccination and the decrease in rates of infectious disease with improved social conditions. In this context investing our limited resources directly into the hospital system can be seen as a sticking plaster approach to health care reform. We may get a far greater return on our investment by using these resources to improve the basic living conditions of populations in deprived areas. This theory will of course require confirmation that intervention will actually lead to improvements in outcomes, there may be other unmeasured confounding factors influencing this association.
In conclusion the major determinants of hospital admission in our cohort were Deprivation Index and population Dependency Ratio. Furthermore there was a synergistic effect between these two factors.
5. Learning points
•
Deprivation Index and Dependency Ratio are important determinants of hospital admission rates.
•
Known outcome predictors such as Acute Illness Severity Score and Chronic Disabling Disease Score do not independently predict hospital admission rates.
•
Interventions to target deprivation may have the potential to reduce hospital admissions.
Conflict of interest
None of the authors have any conflict of interest to declare.
Are hospital services used differently in deprived areas? Evidence to identify commissioning challenges Oxford: Centre for Health Service Economics & Organisation.
Variations and inter-relationship in outcome from emergency admissions in England: a retrospective analysis of Hospital Episode Statistics from 2005–2010.
Association of population and primary healthcare factors with hospital admission rates for chronic obstructive pulmonary disease in England: national cross-sectional study.
00295-2&id=gr1.jpg){kind=link}
00295-2&id=gr2.jpg){kind=link}
00295-2&id=gr3.jpg){kind=link}
00295-2&id=gr4.jpg){kind=link}