Computer Methods and Programs in Biomedicine 56 (1998) 205–210

A comprehensive register for diabetic outpatients: experiencewith desktop computing from 1987–1996

William Kelly a,*, Rudolf Bilous a, Gordon Murray b

a Diabetes Care Centre, Middlesbrough General Hospital, Middlesbrough TS5 5AZ, UKb School of Computing and Mathematics, Uni6ersity of Teesside, Middlesbrough, UK

Received 28 January 1997; received in revised form 7 November 1997; accepted 23 December 1997

Abstract

Diabetes mellitus affects all parts of the body, and all aspects of life for the patient. To manage patients effectively,we need a record of who has diabetes, where they are treated, by whom, and with what results. In 1985 we deviseda standardised data collection form in our clinic and in 1987 we established a comprehensive register of patientstreated in our clinics. Our Diabetes Care Centre serves a population of 287460 based in an industrial area with highlevels of social and economic deprivation. The database includes demographic details, risk factors including smokingand alcohol intake, duration of diabetes, type of treatment, biochemical results including blood glucose, glycosylatedhaemoglobin, details of blood lipids and renal function. We also record details of diabetic eye disease, foot disease,nerve pathway disease, cardiovascular disease, and skin complications. We record the type of treatment including diet,tablets, and insulin and we audit the incidence of diabetic complications including ketoacidosis, amputation, andblindness due to diabetes. From 1994 onwards we compiled a district register of patients including those not seen byus. Our computerised database enables us to monitor our clinics, audit our results, pursue research, set targets, andfacilitates contracting for diabetes care with our purchasers. © 1998 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Diabetes mellitus; Diabetes registers; Desktop computing

1. Introduction

Diabetes mellitus is a chronic medical conditioncharacterised by raised blood glucose concentra-tions. Treatment is with diet, hypoglycaemic

medicines, and insulin injections. There is con-vincing evidence that intensive treatment of dia-betes, with a target of normalising blood glucoselevels, will diminish the risks of diabetic complica-tions [1]. These complications include retinopa-thy—which can lead to blindness, kidneydisease—which can lead to renal failure, andnerve pathway disease—which can lead to impo-

* Corresponding author. Tel.: +44 1642 850850, ext 5722;fax: +44 1642 854327.

0169-2607/98/$19.00 © 1998 Elsevier Science Ireland Ltd. All rights reserved.

PII S0169-2607(98)00026-1

W. Kelly et al. / Computer Methods and Programs in Biomedicine 56 (1998) 205–210 207

Fig. 1. Diabetes Register Form 1996: present or yes, 1; absent or no, —; no details, blank or X; inactive/treated, 2.

W. Kelly et al. / Computer Methods and Programs in Biomedicine 56 (1998) 205–210208

Fig. 2. Entity relationship diagram for diabetes database: �, obligatory (must have); �, non-obligatory(may have); —, one to onerelationship; —B , one to many relationship.

duced to allow quick and easy entry and retrievalof patient information. The system in the form ofan entity relationship diagram is presented (Fig.2). The data base has six tables. A table storesinformation of each patient’s demographics, e.g.name, address, etc. The latest visit table stores thelatest or only visit information. The GP tablestores GP details and is linked directly to thedemographics table. Two other tables are usedwhen patients die or are discharged from ourpatient care. All tables are linked via an internalidentification number. Automatic print functionsinclude single records on preprinted forms, andcomplex summaries, for example the annual sum-mary. Smartware II although a relatively old database management system has a powerful ability tomanipulate the data to suit our requirements. On

January 1st each year we archive previous recordsand commence a new file, so only patients actu-ally seen in each calendar year are included.

5. Status report

In 1987 1089 patients made a total of 2509visits. There were 54.6% male and 227 total newreferrals. Twenty-six percent of patients smoked,average blood glucose was 11.2 mmol/l, and gly-cosylated haemoglobin corrected to the DiabetesControl and Complications Trial (DCCT) refer-ence standard [1] was 9.45%. Twelve percent weretreated by diet alone, 33% by tablets and 55% byinsulin. The system demonstrated a considerableincrease in work, and in 1995, 2163 patients made

W. Kelly et al. / Computer Methods and Programs in Biomedicine 56 (1998) 205–210 209

4071 visits. There were 56.2% male and 263 pa-tients were new referrals. Some patients (6.6%)were from ethnic minorities, mainly from Asia.Age at diagnosis averaged 46.8 years and durationof diabetes mellitus 10.9 years. Comparing 1987with 1995, smoking had decreased to 19.9% (PB0.01) but blood glucose on average remained vir-tually the same at 11.3 mmol/l. Glycosylatedhaemoglobin had improved to 7.9%, and choles-terol was average for a non-diabetic population at6.1 mmol/l. In 1995 11% were treated by dietalone, 44% by tablets and 45% by insulin. Homeblood glucose testing has gradually increased to75%, urine testing alone is done by 17%. Rankingpatients by quintiles of social deprivation by cen-sus data showed a trend towards obesity, andincreased prevalence of smoking in the most de-prived patients (PB0.01) [3,4]. From 1994 wecompiled a district wide register of patients in-cluding for the first time those looked after exclu-sively by their GP [5]. We compared 1887 patientsseen in our hospital clinics in 1994 with 1776patients seen in general practice; smoking preva-lences were almost equal (18.3 vs. 20.9%) butpatients attending the hospital were younger mean52.8 versus 65.4 years with significantly moreproteinuria 24.5 versus 8.7%, with longer durationof diabetes 10.6 versus 7.5 years, and greaterpercentage treated by insulin (43.9 vs. 3.8%). Weinformed those general practices where an appar-ently low prevalence of diabetes occurs; 2 yearslater in 1996 we had identified 5203 patients withdiabetes, giving a prevalence of 1.8% of 287460total population (1.3% in 1994 predicted at 2% for1997). We were able to target and quantify pa-tients and populations at high risk, e.g. for 1852hospital patients with known body mass index(BMI in Kg/m2) 33.9% were obese (BMI\30),only 2.6% were underweight (BMIB20) with63.4% intermediate (BMI 20–30). We can identifywhen data has not been obtained, e.g. for 2271hospital patients, creatinine was known for 1882,or 82.9%: 377 of these 1882 patients (20%) hadcreatinine\120 mmol/l, indicating impaired renalfunction.

We have a liaison nursing sister who has visitedeach general practice to confirm the acceptabilityof the data entry form. We have encouraged

criticism and comment from recipients in generalpractice, so the document has been modified toinclude additional fields, or to clarify ambiguity.

Data entry is done by an experienced assis-tance, who also compiles databases from en-docrine, lipid, diabetic foot, pregnancy andgeneral medical clinics. We have not attempted toidentify those costs applicable specifically to theDiabetes Register. Like others in the real world ofbusy doctors working in overbooked clinics wehave been unable to ascertain 100% of data forevery field. However, we can easily identify whichfields are incomplete for any patient, and whichdoctor was responsible for seeing the patient: Inour team meetings we discuss the rank order forclinic doctors of completeness of data collection,which usually improves when the exercise is re-peated. Where data, e.g. glucose, glycosylatedhaemoglobin, are missing from the data entryform, we interrogate the computer in the Chemi-cal Pathology Department of the hospital. Foreach year since 1987 (except 1988) we havearchived the data to a master file; using this wecan follow individuals or cohorts of patients, forexample to study patients smoking or withretinopathy.

6. Lessons learnt

There are certain problems and benefits in de-veloping a register of diabetic patients [6,7] andapplying medical informatics in diabetes care [8–10]. At first sight, our data entry form (Fig. 1)looks difficult to complete. However, after a fewminutes instruction doctors in training and nursesfind it easy to use and rapidly appreciate itsbenefits and the way in which data are collected ina structured manner. Where information wasdifficult to obtain, or the definition was ambigu-ous, then we have modified the form accordingly.At each clinic up to 10% of the patient’s clinicalrecords can be missing, and we are able to use ourdata base to obtain a print out of the latestinformation from a previous visit. We have beenable to monitor changes in the number and casemix of patients attending our clinic which hasbeen valuable for management purposes in order

W. Kelly et al. / Computer Methods and Programs in Biomedicine 56 (1998) 205–210210

to inform the purchasing and provision of dia-betes services in our health district [11,12]. Ourmoto has always been ‘do the simple things well’.By using our system we have been able to pursueresearch, and audit our results, and to see whichpatients are not having regular check ups, and totarget patients especially at risk of diabetic com-plications [13–17].

7. Future plans

We plan to network our system using DOS as aplatform running Novell Netware 4.1 linked tothe main hospital Unix network. We are exploringways of keeping our system live so that data canbe retrieved, and possibly entered, by all thosewho have a need to do this. We will need to makesure that the data are restricted to those who havepermission to obtain it and that we conform withthe Data Protection Act. We are exploring waysof transferring the data electronically to improvethe speed of communication, and to prevent du-plication of effort. We will ensure that each pa-tient has had the screening and treatment thatcomply with the recommendations from theBritish Diabetic Association and the Departmentof Health [18].

Acknowledgements

We thank Rashad Mahmood for devising ourinitial Diabetes Register. We thank the NorthernRegional Health Authority, Teesside Health JointPurchasing Authority, and South Tees AcuteHospitals NHS Trust for financial support.

References

[1] The Diabetes Control and Complications Trial ResearchGroup, The effect of intensive treatment of diabetes onthe development and progression of long-term complica-tions in IDDM, N. Eng. J. Med 329 (1993) 977–985.

[2] W.F. Kelly, R. Mahmood, M. Arthur, M. Kelly, E. Scott,Inner city deprivation and diabetes control: a computerdatabase, in: B. Richards (Ed.), Current Perspectives in

Healthcare Computing, BJHC, Weybridge, 1993, pp.127–134.

[3] N. Unwin, D. Binns, K. Elliott, W.F. Kelly, The relation-ship between cardiovascular risk factors and socio-eco-nomic status in people with diabetes, Diab. Med. 13(1996) 72–79.

[4] W.F. Kelly, R. Mahmood, M. Kelly, S. Turner, K.Elliott, Geographical mapping of diabetic patients fromthe deprived inner city shows less insulin therapy andmore hyperglycaemia, Diab. Med. 11 (1994) 344–348.

[5] W. Kelly, P. Sherriff, R. Ditta, K. Hogg, V. Connolly,Selective referral of patients to hospital diabetes services:Identification using a district register, Diab. Med. 12(1995) S34 (P30).

[6] S.D. Burnett, C.M. Woolf, J.S. Yudkin, Developing adiabetic register, Br. Med. J. 302 (1992) 627–630.

[7] P. Sherriff, W.F. Kelly, R. Ditta, K.P. Hogg, Problemsand benefits of compiling a district diabetes register,Diab. Med. 12 (1995) S9 (P28).

[8] E.R. Carson, U. Fisher, E. Salzsieder (Eds.), Models andcomputers in diabetes research and diabetes care, Com-put. Methods Programs Biomed. 32 (1990) 171–350.

[9] P.G. Kopelman, J.C. Michell, A.J. Sanderson, DIA-MOND: a computerised system for the management andevaluation of district-wide diabetes care, Diab. Med. 12(1995) 83–87.

[10] J.R. Flack, P.D. Home, An approach to the use of theDO IT study group for supporting the optimal implemen-tation of information systems in diabetes care, Diab.Med. 10 (1993) 568–573.

[11] V. Connolly, N. Unwin, R. Ditta, E. Sayer, R. Bilous, W.Kelly, Age-specific prevalence of known diabetes in atypical British district: evidence of an increase especiallyin the elderly, Diab. Med. 12 (1995) S31 (P17).

[12] V. Connolly, N. Unwin, E. Sayer, R. Bilous, W. Kelly,Increased prevalence of NIDDM but not IDDM in areasof social deprivation, Diab. Med. 13 (1996) S56 (P156).

[13] S. Hurel, W. Kelly, M. Swainston, Which factors arelinked with progression to worsening diabetic retinopa-thy? A cohort study of 389 patients over 4 years, Eur. J.Ophthalmol. 3 (1993) 158.

[14] W.F. Kelly, V. Connolly, G.J. Murray, Anti-smokingadvice and reduction in smoking prevalence in patientswith diabetes 1987–1995, Diab. Med. 13 (1996) S58(P165).

[15] W.F. Kelly, R. Mahmood, M. Kelly, S. Turner, K.Elliott, Influence of social deprivation on illness in dia-betic patients, Br. Med. J. 307 (1993) 1115–1116.

[16] W.F. Kelly, J. Nicholas, J. Adams, R. Mahmood, Necro-biosis Lipoidica Diabeticorum: association with back-ground retinopathy, smoking and proteinuria—a casecontrol study, Diab. Med. 10 (1993) 725–728.

[17] S. Hurel, M. Arthur, M. Swainston, W.F. Kelly, DiabeticKetoacidosis—a case control study showing the impor-tance of patient education, Diab. Med. 10 (1993) S32 33.

[18] G. Higginson (and committee), Standards of ClinicalCare for People with Diabetes, HMSO, London, 1994.

.

W. Kelly et al. / Computer Methods and Programs in Biomedicine 56 (1998) 205–210206

tence and loss of sensation in the limbs. Diabetesmellitus greatly increases the risk of vascular dis-ease including myocardial infarction, cerebrovas-cular accidents, and peripheral vascular disease.Diabetes mellitus is the commonest cause ofblindness between the ages of 20 and 65 in theUK, and is the second commonest cause of endstage renal disease requiring dialysis and trans-plantation in western Europe. It is estimated that�2% of the population of the UK have diabetes,although not all patients have been identified, andthey account for �8% of the total costs of healthcare. Maternal diabetes increases the risk of stillbirth or congenital malformation. Patients withdiabetes have restrictions on occupation, healthinsurance, and those on insulin injections need totake special precautions, for example when travel-ling long distance or altering their diet. For all ofthese reasons it is important to have a databasewhich shows who has diabetes, how, where andby whom they are treated, and the clinical out-comes of therapy. Therefore, we have designed,and gradually modified, a database for routineuse in our diabetic outpatient clinic.

2. Background

Until 1985 the record of diabetic outpatientclinic visits in Middlesbrough was kept on con-ventional lines, that is a paper based system inwhich patients had a separate file which wassupposed to be available at each outpatient visit,and was to be sent to the ward in the event ofadmission to hospital. The records were handwritten, and it was frequently difficult to tell whathad, or had not, been done and whether thisinformation had been communicated to all whohad a need to know including the general practi-tioner, and other hospital consultants. Therefore,in 1985 we commenced a standardised data collec-tion form in our clinic, and this was modified in1987 when we established a comprehensive com-puter register. This was based on a 286 seriesdesktop computer using Smartware II (Informix).We devised a triplicate form [2] the top copygoing to the general practitioner (GP), the secondgoing into the case records and the third copy

went to our data entry clerk. The database couldthen be interrogated to obtain clinical details inthe event of missing case records, and it waspossible to provide annual summaries to showchanging work loads and to monitor the morbid-ity associated with diabetes mellitus in our pa-tients. During the last 10 years we have greatlyincreased the number of fields collected, pro-gressed to 386, then 486 series desktop computers,and data is analysed using the Statistical Packagefor the Social Sciences (SPSS).

3. Design considerations

The data entry form comprises a single sidedA4 sized (21×30 cm) triplicated NCR sheet. Anoveriding consideration has been to maintain theform as a single sided sheet and the emphasis hasbeen on collecting data which was clearly categor-ical, or could be quantified, e.g. blood glucose,glycosylated haemaglobin, creatinine, and choles-terol concentrations. A copy of the latest versionof our data entry form is shown (Fig. 1). Separatesections are provided for demographic details, eyeand foot examinations, vascular disease, skin,pregnancy status, results of blood tests, treatment,and clinical audit. We record the ethnic origin ofthe patient, whether the relatives are affected, andwhether the patient has shared care with theirgeneral practitioner. The current status and nextappointment are also indicated.

4. System description

The diabetes system operates from an IBMcompatible PC, MS-DOS platform. The PC iscurrently a 486DX/2 running at 66Mhz. It has amoderately sized IDE hard disk drive (480MBHDD), with 8 Mb of RAM (SIMMS). For severalyears this system has been running under In-formix’s Smartware II (version 3.10) integratedpackage. Based on the data base module ofSmartware II, the system enlists the use of anumber of relational tables. These tables representthe majority of the information collected from thediabetic forms. A user interface has been pro-