CLINICAL PRACTICE

Hospital Utilization as a Function ofSocial Deprivation: Diabetes vs Non-DiabetesChristopher Ll. Morgan1, Craig J. Currie2, John R. Peters1

1Department of Medicine, University Hospital of Wales, Heath Park,Cardiff, Wales, UK,2Department of Public Health Medicine, Bro Taf Health Authority,Temple of Peace and Health, Cathays Park, Cardiff, Wales, UK

We tested the hypothesis that a relationship between ill health and deprivation exists forpatients with diabetes, distinct from that experienced by the non-diabetic population. Agestandardized admission and appointment rates and proportion of total activity for patientswith and without diabetes were determined by electoral ward and correlated with theTownsend index of social deprivation for the health district of South Glamorgan (population408 000). Both diabetic (r = 0.78, p , 0.001) and non-diabetic (r = 0.74, p , 0.001) in-patient admissions were positively correlated with social deprivation. This relationshipalso existed for attended out-patient appointments (r = 0.67, p , 0.001 and r = 0.45,p , 0.01, respectively). The proportion of diabetic to non-diabetic admissions by wardalso showed a positive correlation for in-patients (r = 0.47, p , 0.001). This remained truefor IDDM (r = 0.23, not significant) and NIDDM (r = 0.62, p , 0.001) diabetes, foradmissions for coronary heart disease (r = 0.50, p , 0.001) and cerebrovascular disease(r = 0.29, p , 0.05), elective admissions (r = 0.30, p . 0.05), and emergency admissions(0.46, p , 0.001). Our results suggest that secondary care utilization is positively correlatedwith social deprivation and that this relationship is stronger in the diabetic population.This may be due to different prevalence rates or increased complications requiring hospitaltreatment in different social circumstances. Further research is required to examine thesefactors more closely. 1997 by John Wiley & Sons, Ltd.

Diabet. Med. 14: 589–594

No of Figures: 3. No of Tables: 3. No of Refs: 21

KEY WORDS Diabetes mellitus Social deprivation Townsend index In-patientadmissions Out-patient appointments

Received 16 October 1996; revised 15 February 1997; accepted 15 February 1997

diabetes and to determine whether this relationship isIntroductiondistinct from that observed for the population in general.

The positive relationship between ill health and socialdeprivation has been well documented1,2 and recent Patients and Methodsstudies suggest that this relationship is also true forpatients with diabetes.3 Cardiovascular disease and its The study was conducted in South Glamorgan, a typical

teaching health district with a population of 408 000associated risk factors are more prevalent among sociallydeprived patients with diabetes than those less and a centre for cardiac, renal, and neurological services.

It is divided administratively into 47 electoral wards.deprived.4,5 Rates of in-patient admissions for patientswith diabetes are also strongly correlated with increasing The overall prevalence of diabetes in South Glamorgan

has been estimated at 1.36 %.8levels of social deprivation.6 It is not clear howeverwhether the relationship between ill health and social All routine in-patient admissions over a 4-year period

(1991–1995) and out-patient appointments over a 3-deprivation experienced by the diabetic population,given the greater morbidity associated with diabetes,7 is year period (1991–1994) were retrospectively coded as

referring to a diabetic or non-diabetic patient. Thesolely a function of that experienced by the populationas a whole. method used for identifying patients with diabetes has

been described in detail elsewhere.7We wished to confirm that a relationship betweensocial deprivation and ill health exists for patients with Data on residents from outside South Glamorgan and

those treated as tertiary referrals were excluded and eachof the remainder ascribed a deprivation score basedupon electoral ward using the Townsend index.9 Electoral* Correspondence to: Dr J.R. Peters, Department of Medicine, University

Hospital of Wales, Heath Park, Cardiff, CF4 4XW, Wales, UK ward was assigned on the basis of postcode. The

589CCC 0742–3071/97/070589–06$17.50 1997 by John Wiley & Sons, Ltd. DIABETIC MEDICINE, 1997; 14: 589–594

CLINICAL PRACTICETownsend index incorporates four variables: proportion between social deprivation and ill health we wished to

determine whether any observed variation was due toof population of working age unemployed, proportionof households without a car, proportion of households different diabetes prevalence rates or different rates of

admissions. In the absence of ward prevalence data, weovercrowded, and proportion of households not owner-occupied. Age-standardized rates of in-patient admissions considered the rate of individuals who were admitted,

as in-patients over the 3-year period. The ratio of theseand out-patient appointments by electoral ward werecalculated for both the diabetic and non-diabetic popu- patients to admissions by ward for both the diabetic and

non-diabetic populations was also calculated.lation using the total resident ward population as thedenominator. Age-standardized in-patient and out-patient

Resultsactivity was also expressed as a proportion of totalactivity for each electoral ward and averaged for the

There were 358 018 in-patient admissions and 831 147five least and five most deprived. The rates and pro-out-patient appointments of which 349 989 (97.8 %) andportions were compared with the Townsend index using787 139 (94.7 %), respectively, were for South Glamorganthe Pearson correlation coefficient.residents whose electoral ward could be deduced byDiabetes type was identified from hospital records andtheir postcode. Patients with diabetes accounted fornot on more stringent criteria. However, using these22 104 (6.3 %) in-patient admissions. After excludingdefinitions we repeated our calculations for insulin-those patients who did not attend their out-patientdependent (Type 1) diabetes mellitus (IDDM) and non-appointments there were 606 669 contacts, 48 681insulin dependent (Type 2) diabetes mellitus (NIDDM)(8.0 %) relating to patients with diabetes.and also for in-patient admissions over 35 years of age

Of diabetic in-patient admissions, 3731 (16.9 %) werewith a primary diagnostic code relating to coronary heartfor patients with IDDM and 13 584 (61.4 %) for patientsdisease (CHD) (ICD 9 codes 401, 410–414, 426, 428)with NIDDM. Diabetic type was not recorded on 4789and cerebrovascular disease (CVD) (ICD 9 codes 434–(21.7 %) records, which were excluded from this sub-436, 438). Analysis was also performed separately foranalysis. There were 19 139 (5.5 %) in-patient admissionselective and emergency admissions.with a primary diagnostic code relating to CHD andTo examine more closely the nature of the relationship6625 (1.9 %) relating to CVD. Of these, 2869 (14.9 %)and 1001 (15.0 %), respectively, were for patients

Table 1. Age standardized rate of inpatient admissions corre- with diabetes.lated to deprivation of electoral ward

Social Deprivation and Diabetic In-patientr p value and Out-patient Contacts

All diabetic in-patients 0.78 ,0.001 Tables 1–3 show the age-standardized rates of in-patientAll non-diabetic in-patients 0.74 ,0.001 admissions, out-patient appointments, and proportion ofInsulin requiring diabetic in-patients 0.47 ,0.01

activity in relation to the Townsend index by electoralNon-insulin requiring diabetic in- 0.83 ,0.001ward. Both overall diabetic (Figure 1) and non-diabeticpatients

Diabetic coronary heart disease in- 0.70 ,0.001 (Figure 2) admission rates displayed a positive correlationpatients with social deprivation (r = 0.78, p , 0.001 and r = 0.74,Non-diabetic coronary heart disease 0.42 ,0.01 p , 0.001, respectively). A correlation was also foundin-patients

between social deprivation and out-patient appointmentDiabetic cerebrovascular disease in- 0.44 ,0.01rates for patients with (r = 0.67, p , 0.001) and withoutpatients

Non-diabetic cerebrovascular disease 0.39 ,0.05 diabetes (r = 0.45, p , 0.01). The proportion of diabeticin-patients to total in-patient activity (Figure 3) showed a positiveDiabetic emergency admissions 0.85 ,0.001

correlation to deprivation (r = 0.47, p , 0.01). DiabeticNon-diabetic emergency admissions 0.87 ,0.001admissions accounted for 5.1 % of total in-patientDiabetic elective admissions 0.30 ,0.05admissions in the five least deprived electoral wardsNon-diabetic elective admissions 0.06 N/Scompared with 7.2 % in the five most deprived electoralwards. A correlation was also found for the proportionof out-patient activity to social deprivation (r = 0.45,

Table 2. Age standardized rate of out-patient attendances p , 0.01) with 6.8 % of appointments due to patientscorrelated to deprivation of electoral ward

with diabetes in the least deprived electoral wardscompared with 8.8 % in the most deprived.r p value

IDDM and NIDDMDiabetic out-patients 0.67 ,0.001Non-diabetic out-patients 0.45 ,0.01

Considering only IDDM there was a weak correlationDiabetic out-patients DNAs 0.31 ,0.05Non-diabetic out-patients DNAs 0.81 ,0.001 (r = 0.47, p , 0.01) for in-patient admissions. NIDDM

however showed a strong correlation (r = 0.83,

590 C.LL. MORGAN ET AL.

CLINICAL PRACTICETable 3. Proportion of diabetic activity to total activity by electoral ward

5 most affluent 5 most deprived r p value(%) (%)

Diabetic in-patients 5.1 7.2 0.47 ,0.01Diabetic out-patients 6.8 8.8 0.45 ,0.01IDDM in-patients 1.0 1.2 0.23 N/SNIDDM in-patients 2.4 4.4 0.62 ,0.001Diabetic coronary heart disease in-patients 13.2 18.7 0.50 ,0.001Diabetic cerebrovascular disease in-patients 9.5 20.8 0.29 ,0.05Diabetic elective admissions 4.1 6.0 0.30 ,0.05Diabetic emergency admissions 7.0 8.9 0.46 ,0.01

Figure 1. Diabetic inpatient age standardized admissions to deprivation of electoral ward

p , 0.001). The separate correlations of IDDM and electoral wards against 18.7 % in the five most deprived.Diabetic admissions as a proportion of total CVDNIDDM as a proportion of total admissions by electoral

ward were also positive (r = 0.23, p . 0.1 and r = 0.62, admissions showed a less significant correlation withdeprivation (r = 0.29, p , 0.05). Patients with diabetesp , 0.001, respectively). IDDM accounted for 1.0 % of

admissions in the five least deprived electoral wards represented 9.5 % of CVD admissions in the leastdeprived electoral wards compared with 20.8 % in thecompared with 1.2 % in the five most deprived. The

corresponding proportions for NIDDM were 2.4 % and most deprived.4.4 %.

Elective and Emergency AdmissionsVascular Disease

Diabetic elective admissions were correlated with socialdeprivation (r = 0.30, p , 0.05) but there was no corre-Vascular disease specific analysis showed a positive

correlation for in-patient admissions and increasing social lation for the non-diabetic population (r = 0.06, p . 0.20).This was reflected in the proportion of diabetic electivedeprivation. The proportion of diabetes to non-diabetes

contacts with a primary ICD code relating to CHD admissions to total elective admissions (r = 0.30,p , 0.05). The average proportion of diabetic electiveincreased with increasing deprivation (r = 0.50,

p , 0.001) with 13.2 % of total CHD admissions relating admissions in the five least deprived wards was 4.1 %compared with 6.0 % in the five most deprived.to patients with diabetes in the five least deprived

591HOSPITAL UTILIZATION, SOCIAL DEPRIVATION AND DIABETES

CLINICAL PRACTICE

Figure 2. Non-diabetic inpatient age standardized admissions to deprivation of electoral ward

Figure 3. Ratio of diabetic activity to total activity by deprivation of electoral ward

592 C.LL. MORGAN ET AL.

CLINICAL PRACTICEIn contrast to the pattern of elective admissions, demonstrate no clear relationship between the ratio of

patients to admissions and social deprivation by wardparticularly in the non-diabetic population, emergencyadmissions showed a strong correlation with social for the diabetic population whereas the number of

individuals admitted is strongly correlated. This woulddeprivation for both diabetic and non-diabetic patients(r = 0.85, p , 0.001 and r = 0.87, p , 0.001, suggest that the cause of increased activity in the more

deprived areas is a higher prevalence of diabetes.respectively). Patients with diabetes accounted for 7.0 %of total in-patient emergency admissions in the five least However this is not conclusive. Differences in risk factors

and accessibility to primary services, for example, maydeprived wards compared with 8.9 % in the five mostdeprived (r = 0.46, p , 0.001). ensure that those from the most affluent areas are

possibly less likely to come into contact with secondarycare services.Non-attendance at Out-patient Clinics

Risk factors for the complications of diabetes (smoking,hypertension, obesity, high glycated haemoglobin, hyper-There was a strong correlation between social deprivationlipidaemia, and blood glucose levels) have been demon-and non-attendance at out-patient clinics for the non-strated to be more prevalent among those from depriveddiabetic population (r = 0.81, p , 0.001) but this wasareas.4,13 Diabetes is less well controlled among thoseweaker for those patients with diabetes (r = 0.31,most deprived,13 where provision and accessibility top , 0.05). There was no significant correlation betweenhealth care is reduced. Diabetic clinics are less commondiabetic non-attendance as a proportion of total non-in deprived areas14 and the uptake of invitations toattendance (r = 0.03, p . 0.20).primary care health promotion sessions for the populationin general has been shown to fall with social class.15

Rate of Individuals and Ratio of Patients Our results suggest that this situation exists in theto Admissions secondary care setting. Non-attendance at out-patient

appointments is correlated with social deprivation andA strong correlation was also found between social is especially strong for the non-diabetic population.deprivation and the rate of individuals admitted in each There is a lack of correlation for the non-diabeticward for both the diabetic and non-diabetic populations population and only moderate correlation for the diabetic(r = 0.83, p , 0.001 and r = 0.68, p , 0.001, population between elective in-patient admissions andrespectively). The proportion of individuals with diabetes social deprivation, in contrast to the strong correlationto those without diabetes also showed a positive corre- with emergency admissions for both groups. This, takenlation (r = 0.70, p , 0.001). The ratio of patients to with our vascular disease specific data, suggests a greateradmissions for the non-diabetic population also displayed prevalence of morbid conditions likely to result ina correlation with social deprivation (r = 0.68, p , 0.001) emergency admissions among the most deprived, butbut there was no significant correlation for patients with it may also be partly due to later presentation ofdiabetes (r = 0.12, p . 0.02). serious illness.

NIDDM showed a stronger correlation with socialdeprivation than IDDM. This confirms the findings ofDiscussionprevious studies.16,17 Due to the number of records whichwere not coded for type of diabetes any conclusions madeThese results suggest that not only is there a linear

relationship between ill health and social deprivation concerning the correlation between social deprivationand diabetic type should be treated with caution,but that this relationship, when measured by utilization

of secondary care services, is stronger for patients with however, we have no reason to believe that the non-coding of diabetic type on patient records is systematicallydiabetes than for the population as a whole. This

remains true when considering sub-analyses of IDDM related to social deprivation or any other factor thatwould cause the proportion of IDDM and NIDDM inand NIDDM, elective and emergency admissions and

vascular disease specific admissions. Previous studies the non-coded records to differ significantly from thosethat were coded.have indicated that increasing levels of social deprivation

are positively correlated to increasing prevalence of We examined two subsets of in-patient admissionrecords with groups of primary diagnostic codes relatingNIDDM and diabetes associated complications.3,10,11 A

recent study showed no correlation between deprivation to CHD and CVD in addition to overall admissions.Consideration of overall admissions alone may proveand self-reported diabetic prevalence but a strong associ-

ation with diabetic eye disease possibly indicating poorer misleading as many of the co-morbidities associated withdiabetes are known to be influenced independently bycontrol, compliance or screening for the deprived popu-

lation.12 deprivation and its associated risk factors. Smoking,obesity, and other risk factors for CHD have been shownThe observation of increased in-patient activity by

social deprivation of electoral ward may be due to to be inversely related to social class18 and CHDrepresents a major cause of morbidity among patientsdifferent prevalence rates of diabetes, different rates of

complications or a combination of both. These results with diabetes. This may distort the apparent excess

593HOSPITAL UTILIZATION, SOCIAL DEPRIVATION AND DIABETES

CLINICAL PRACTICE2. Macintyre S. Understanding the social patterning of health:relationship between diabetes and deprivation over and

the role of the social sciences. J Public Health Med 1994;above that of the general population. By considering16: 53–59.CHD and CVD as separate subsets within the diabetic 3. Kelly WF, Mahmood R, Kelly MJ, Turner S, Elliott K.

and non-diabetic populations we have controlled for this Influence of social deprivation on illness in diabeticand demonstrated that the relationship between the patients. Br Med J 1993; 307: 1115–1116.

4. Unwin N, Binns D, Elliott K, Kelly WF. The relationshipincreasing proportion of diabetic to non-diabetic morbidbetween cardiovascular risk factors and socio-economicevents with increasing deprivation, although reduced,status in people with diabetes. Diabetic Med 1995; 13:remains true and independent of other deprivation72–79.associated risk factors. 5. Connolly VM, Kesson CM. Socioeconomic status and

Data to allow classification of social deprivation on clustering of cardiovascular disease risk factors in diabetican individual basis were not available and as a result patients. Diabetes Care 1996; 19: 419–422.

6. Caddick SL, McKinnon M, Payne N, Ward TJ, Thornton-scores had to be ascribed by electoral ward. AlthoughJones H, Kells J, et al. Hospital admissions and socialno aggregated measure of deprivation is ideal, thedeprivation of patients with diabetes mellitus. DiabeticTownsend index was designed specifically for use withinMed 1994; 11: 344–348.the health arena and has been shown to compare closely 7. Currie CJ, Williams DR, Peters JR. Patterns of in and out-

to the concept of material disadvantage.19 The allocation patient activity for diabetes: a district survey. Diabeticof ward on the basis of postcode may have led to a bias Med 1996; 13: 273–280.

8. Butler C, Peters JR, Stott N. Glycated haemoglobin andif there was any differential non-coding of postcode butmetabolic control of diabetes mellitus: external versusthe high proportion of records which could be codedlocally established clinical targets. Br Med J 1995; 310:minimizes any effect that may have arisen.748–788.A partial explanation for these results may be a ‘drift’ 9. Townsend P, Phillimore P, Beattie A. Health and Depri-

theory, whereby the morbidities associated with diabetes vation: Inequality and the North. London: Croom Helm,are more severe than those in general population. 1988.This may result in patients with diabetes being less 10. Meadows P. Variation of diabetes mellitus prevalence in

general practice and its relation to deprivation. Diabeticeconomically active and migrating into areas of increasedMed 1995; 12: 696–700.social deprivation. A previous study by Barker et al.

11. Chaturvedi N, Stephenson JM, Fuller JH. The relationshipdisplayed a relationship between diabetes incidence and between socioeconomic status and diabetes control andthe social deprivation status of towns but could not complications in the EURODIAB IDDM complicationsreconcile this with incidence at an individual level study. Diabetes Care 1996; 19: 423–430.measured by social class.16 This could be partly explained 12. Eachus J, Williams M, Chan P, Davey Smith G, Grainge

M, Donovan J, Frankel S. Deprivation and cause specificby the fact that while such migrants might maintain theirmorbidity: evidence from the Somerset and Avon surveysocial class, by any aggregate index based upon residence,of health. Br Med J 1996; 312: 287–292.they will be classified as socially deprived.

13. Kelly WF, Mahmood R, Turner S, Elliott K. GeographicalEthnicity, which has been shown to be correlated with mapping of diabetic patients from the deprived inner city

areas of social deprivation, is a confounding factor which shows less insulin therapy and more hyperglycaemia.we could not control in this study. South Glamorgan Diabetic Med 1994; 13: 72–79.

14. Leese B, Bosanquet N. Change in general practice andhas an overall non-white population of 4.8 % which isits effects on service provision in areas with differentmore likely to live in areas of social deprivation.20 Thesocioeconomic characteristics. Br Med J 1995; 311:increased prevalence of diabetes and its complications 546–550.

among Afro-Caribbeans and those from the Indian sub- 15. Waller D, Agass M, Mant D, Coulter A, Fuller A, Jonescontinent has been reported repeatedly since its original L. Health checks in general practice: another example ofdescription.21 The different complication patterns which inverse care? Br Med J 1990; 300: 1115–1118.

16. Barker DJP, Gardner MJ, Power C. Incidence of diabeteshave been reported for the non-white population distortsamongst people aged 18–50 years in nine British towns:any analysis considering activity rates and this requiresa collaborative study. Diabetologia 1982; 22: 421–425.further study.

17. Connolly V, Unwin N, Sayer E, Bilous R, Kelly WF.Regardless of the effect of ethnicity, this study provides Increased prevalence of NIDDM but not IDDM in areas

evidence that there is a need for greater allocation of of social deprivation. Diabetic Med 1995; 12: S56.resources to areas of increasing social deprivation in 18. Pocock SJ, Shaper AG, Cook DG, Phillips AN, Walker

M. Social class differences in ischaemic heart disease inorder to reduce the morbidity caused by diabetes. FutureBritish men. Lancet 1987; (2): 197–201.research is needed to determine the nature of the cause

19. Morris R, Carstairs V. Which deprivation? A comparisonof this morbidity.of selected deprivation indexes. J Public Health Med1991; 13: 318–326.

20. Kennedy DS. South Glamorgan Health and Social CareReferencesProfile. Cardiff: South Glamorgan Health Authority, 1993.

21. Mather HM, Keen H. The Southall diabetes survey:1. Black D, Morris JN, Smith C, Townsend P. Report ofprevalence of known diabetes in Asians and Europeans.the Working Group on Inequalities in Health. London:

DHSS, 1980. Br Med J 1985; 291: 1081–1084.

594 C.LL. MORGAN ET AL.