(excluding patients who were attending hospital eye clinics at the outset of study)
Control Prom pted P
At baseline; Number of patients - 70 np - 74
Number of patients with cataract/extraction Number of patients with non STR
7 (10%) 1 (1%) 4 (5%) 2 (3%) NS NS Process of care nj. - 70 np = 74
Number of patients who did not attend
Mean num ber of eye exam inations per patient per year n^ - sa np - 72
Number of patients referred to hospital eye clinic
12 (17%) 0.9 (0.8) 11 (19%) 2 (3%) 1.1 (0.8) 7 (10%) 0.008 NS NS O utcom e n^ »> sa np = 72
Number of patients with new cataract Number of patients developing non STR Number of patients with new STR
3 (5%) 2 (4%) 5 (9%) 29 (40%) 2 (3%) 2 (3%) <0.001 NS NS Note: S IR = sight threatening retinopathy.
All values listed, other than baseline are taken from asse ssm e n ts nearest to the en d of the study and unless otherwise stated are given a s m ean (SD).
n^j = num ber in control group; Op = num ber in prom pted group; NS = not significant at 5% level. Statistical te sts used: two tailed t-test for duration, Mann-Whitney te s t for rates (adjusted for ties)
test for proportions (with continuity correction)
During the study period, 12 controls and 2 prompted patients did not attend either hospital outpatients or optometry screening (p=0.008). After randomisation, the prompted group received on average 1.1 eye examinations per year compared w i t h 0.9 in the control group. There
was no significant difference between the two groups in the number of patients referred to hospital eye clinics. The number of cataracts newly recorded by optometrists in the prompted group vastly exceeded that recorded by doctors in the hospital clinic group (c: 3 v. p: 29; p<0.001), though the study incidence of newly recorded retinopathy did not differ significantly between the two groups.
Discussion
These results show that a prompting system of structured diabetic care can support appropriate medical care comparable to that provided in a hospital diabetic clinic in the case of non insulin treated patients registered w ith small inner city general practices
in inner London. Prompted care in Islington resulted in a
significantly lower lost to follow-up rate than that achieved in the hospital diabetic clinic.
Professional and patient compliance proved high; the lower default rate in the prompted group is especially important because loss to follow-up carries an increased risk of the onset of diabetic
complications, particularly in non insulin treated patients
(Hammersley et al 1985).
Prompted care achieved six monthly doctor review together w i t h high levels of specific diabetes assessments w i t h more frequent recordings of weight, blood pressure and foot inspection in the prompted group than in controls. All prompted GP reviews were performed in the context of results from recent blood glucose, HbAj and albuminuria estimations. This level of assessment compares favourably w i t h the most comprehensive levels of care reported from hospital clinics, or from GP mini-clinic care (Porter 1979, Yudkin et al 1980, Williams et al 1989, Kemple 1991, Parnell 1993). The system clearly allowed for easy referral of patients to the hospital clinic if deemed necessary by the GP. However, it is important to note that 40% (^^/s2 see Table 6) of prompted patients who were reviewed in a hospital diabetic clinic at some point during the study found their way there without GP referral, though a quarter of
these were patients who changed their minds about accepting prompted care.
Differences in the knowledge and skills of health carers in their very different settings are likely to result in differences in quality of care. While it is recognised that process of care measures are an imperfect surrogate for the standard of patient care, objective measures of medical outcome showed no strong evidence of poorer care in one group than another. There was no evidence, for example, of deterioration in glycaemic control between the two groups, and rates of admission to hospital and mortality were both comparable.
Some of the clinical outcome measures consisted of records of observations performed in routine care settings on the part of a wide variety of doctors not trained to minimise inter- and intra observer variability. Though the proportion of patients recorded as showing ne w onset of lower limb ischaemia was greater in the prompted group, this could be a result of poor GP skills in detecting foot pulses. On the other hand, this result may not have been due to poor examination skills but could have been the result of a higher level of lower limb ischaemia at baseline documented in the prompted group shown in Table 3 (Osmundson et al 1990). Similarly, the rise in mean diastolic blood pressure in the hospital controls, together w i t h a small fall in the prompted group at the end of the study are of questionable significance in v i e w of the likely observer error in these measurements.
The responsibility for retinal screening lay w i t h optometrists in the case of patients not already under the care of a hospital eye clinic at entry to the study. After allowing for a higher non attendance rate in hospital controls, the process of care was comparable in the two groups. Whilst acknowledging there to be no 'gold standard* here, there was also no difference between the two groups in the onset of retinopathy during the study. The m uch higher detection rate of cataract in the prompted group probably reflects the diligence of optometrists in noting these defects compared to
less rigorous criteria used by hospital clinic doctors.
The next chapter looks at the acceptability of this system of care to patients, GPs and optometrists by examining their replies to q u e s t i onnaires.