Improving the health behaviours of elderly people: randomised controlled trial of a general practice education programme

General practitioners

Our project received the approval of the University of Melbourne human ethics committee. We assigned a number to 398 general practitioners from a list of a regional grouping in metropolitan Melbourne, Australia, then randomly selected 193 of these to be recruited by telephone. Eligibility criteria were: working more than 12 hours per week, not planning to move or retire in the next 2 years, one general practitioner per practice site, and no computerised recall system for influenza vaccination. We excluded 6 general practitioners (3%) who were uncontactable, 25 (13%) who had either moved practice or died, 28 (15%) whose partners were already enrolled in our trial, 25 (13%) who worked less than 12 hours per week, 7 (4%) who were retiring, 13 (7%) who had either no elderly patients or patients who did not speak English, and 7 (4%) who had computerised recall systems.

Overall, 42 of 82 eligible general practitioners (6 females and 36 males) were enrolled (51% participation rate), of whom 17 (40%) were the sole doctor in their practice. The 40 general practitioners who declined to participate (50% returned a survey) had been at their current practice a shorter time (t=2.03; P<0.05) and were less likely to charge their patients (χ²=4.57, P=0.03) than those who participated.

Patients

After informed consent of the participating general practitioners, all practice records were counted (average 6657, range 938-66 000 records). On the basis of patient eligibility criteria for the trial (≥65 years, English speaking, community dwelling, attended the practice in the past 18 months, attended the enrolled general practitioner for three of the past five consultations) and a random number table, we selected and viewed an average of 397 (range 50-2000) records per practice, and we identified 10 elderly patients per participating general practitioner. Overall, 267 patients agreed to participate in the trial when invited to do so by post (participation rate 64%). Patient non-participants (60% (92 of 153) contacted by telephone) were more likely to be dependent for transportation (P=0.003) and shopping (P<0.0001) but were otherwise indistinguishable from participants After 1 year, 34 patients (13%) were not followed up (see website).

Randomisation and blinding

An independent research assistant at a distant site used computer randomisation to allocate general practitioners to intervention or control group and this was concealed until the intervention began. Interviewers evaluating outcomes were blinded to the intervention group of patients and general practitioners at all times, and patient's were unaware of the group allocation of their general practitioner.

Outcomes

Patient outcomes were evaluated by trained interviewers during home visits at baseline from November 1995 to February 1996 and at 1 year follow up from December 1996 to April 1997. Outcomes were (a) patients' recall of discussions with their general practitioner; (b) patients' self reported evaluation of physical activity on the basis of questions used in previous surveys 8 9; (c) reports of frequency and duration of activity episodes in the previous fortnight (walking, sports and exercises, gardening, housework and home maintenance), which were multiplied to obtain total minutes per fortnight; (d) extent of walking the day before (walking yesterday) using a previously validated question10; (e) frequency of social contact from asking how often patients did something they really enjoyed; (f) validated measures for psychological wellbeing (psychological subscale of the sickness impact profile)11 and functional status (human activities profile)12; patients' self rated health by asking if, in general, they rated their health as excellent, very good, good, fair, or poor, and by asking how they would rate their health in general now compared with 12 months earlier; (g) total number of drugs taken: psychotropics, non-steroidal anti-inflammatories, and analgesics (coded by NK) as viewed and recorded by interviewers; and (h) patients' self report of influenza vaccination status.

Intervention

General practitioners undertook an educational programme, comprising 5 stages, which began within 2 weeks of patient enrolment and spanned 2-3 months (box) All stages covered areas of social and physical activity and prescribing and vaccination practices for elderly patients. Each general practitioner in the intervention group undertook some or all of the five stages. They were expected to incorporate the intervention into their daily practice and to pass on health promotion advice to patients as appropriate.

Analysis

We used STATA to perform an intention to treat analysis, maintaining patients in their original groups regardless of completion of trial. For those not completing the trial, a gradual decline in outcomes would be expected. We avoided an overestimation of the intervention effect by choosing the more conservative estimate of “no change” for the outcome of participants who had not completed the trial at follow up. We adjusted for the effect of clustered design with a cross sectional time series iterative generalised least squares regression.13 For dichotomous outcomes, follow up status was regressed on status at baseline and intervention group status.14 For continuous measures, we used the change over time as the unit of analysis. The robust option was used to allow for non-normally distributed data where appropriate.15 We adjusted for factors unevenly distributed between the groups. Gains in physical activity were converted to estimated energy expenditure and compared to the Harvard alumni data16 to estimate possible reductions in mortality.

Physical activity

Baseline characteristics of the sample (table 1) were evenly distributed between the groups with the exception of practice billing style (whether the doctor billed the National Health Scheme directly for payment, termed “bulk billing,” or charged the patient). After the trial period, 39 patients (32%) in the intervention group and 21 (19%) in the control group remembered discussing exercise with their general practitioner (P=0.043). Overall, 24 entries for discussion of physical activity appeared in the intervention group records. Nine of these patients recalled the discussion and a further 29 recalled a discussion that had not been noted by the doctor.

Intention to treat analysis showed positive effects of the intervention on patient walking, frequency of pleasurable activities, and self rated health (table 2). On average, the amount of weekly walking in the intervention group was 44 minutes more than that in the control group (95% confidence interval 7 to 168; P<0.032). Reanalysis of variables for physical activity on the basis of the number of people who increased their walking by discrete amounts (on the basis of quintiles and tertiles) also showed an intervention effect (table 2). Correlation between self reported walking in the past fortnight and yesterday was high at both baseline (Pearsons coefficient 0.66, P<0.0001) and after change (0.38, P<0.0001). Correlation of walking with functional status was also significant (0.22, P<0.001 and 0.31, P<0.001 for walking in the previous fortnight v walking yesterday respectively).

At follow up, 12 more patients (10%) in the intervention group had increased their walking and 11 more patients (9%) in the intervention group had not decreased their walking compared with the control group (table 3). The frequency of pleasurable activities and perceived change in self rated health was increased in 20% more of patients in the intervention group than in the control group.

The intervention did not affect functional status, psychological wellbeing, immunisation rate, or total number of drugs used (table 2). Overall, immunisation rate increased from 66% at baseline to 73% at follow up Use of psychotropic drugs including benzodiazepines, antidepressants and major tranquillisers, non-steroidal anti-inflammatories, and analgesics showed no differential changes between the groups. When sedative hypnotic drugs were analysed separately, seven more people in the intervention group started taking sleeping tablets compared with one in the control group (P<0.001). On further inquiry, four of the seven patients in the intervention group initiating sleeping tablets consumed one quarter to one half a tablet fortnightly or weekly, and in two patients an operation and a death in the family had prompted use of sleeping tablets.

The average level of walking at baseline was 1.25 hours per week Assuming a value of 4.5 for metabolic equivalent of work for brisk walking17 and an average body weight of 70 kg, energy expenditure on walking was 393.75 kcal/week, similar to the baseline rate in the Harvard alumni study.16 The intervention group increased walking by an average of 0.73 hours (44 minutes) per week and expended an estimated additional 300 kcal/week. Paffenberger's second category of 500-999 kcal/week16 showed a relative risk of 0.78 from all cause mortality when compared with the baseline group. It is said that the Harvard alumni data applies to females,18 therefore there may be a 22% reduction in mortality associated with the increase in walking observed in our study.

Limitations

Generalisability of our findings is not assured although response rates were similar to other studies Participating general practitioners had been at their current practice site longer and were less likely to bill their patients in bulk than non-participants. Participating patients were more functionally able than non-participants. Reproducability of this result awaits further study.

Variability in the change in activity and width of confidence intervals around positive results suggests caution in interpretation. The walking variables, however, showed a consistent intervention effect on quintiles and tertiles and were observed with impacts on recall of discussions of exercise (however unreliable), self rated health, and frequency of social activities Additionally, the question on walking yesterday that was previously validated against a pedometer10 was highly correlated with other self report walking variables and functional status. This lends some validity to the result.

The estimations of reduction in mortality from this trial may be based on incorrect assumptions. The Harvard alumni group comprised men from a different socioeconomic background than our trial group. Similar reductions in mortality, however, have been associated with similar levels of walking in groups more closely resembling a primary care group.22

General practice intervention

To the authors' knowledge this trial is unique in that general practitioners and elderly patients were randomly selected, the general practitioner was randomised to receive an educational programme, and outcomes were evaluated at the patient level. One trial reported favourable changes in biochemical variables but not exercise behaviour23 and others have not used primary care as the setting.24 25

In our study, walking but not other activities increased. Success of activity interventions was more likely if exercise was enjoyable, of moderate intensity, and not related to attending a facility.26 This programme may have been less expensive or complex than other interventions tested in all ages.27 28 Strategies to aid dissemination need to improve doctors' participation rate.

Conclusion

Education and clinical practice audit for general practitioners improved health outcomes in elderly patients. Although modest, the improvements in physical activity had public health significance and resulted from a comparatively inexpensive intervention.