The impact on general practitioners of the changing balance of care for elderly people living in institutions

The sample

The Office of Population Censuses and Surveys sampled one in 13 establishments (n=1408) possibly providing care for disabled people in Great Britain. Of the 892 establishments eligible for inclusion, 595 were randomly selected for the study. “Permanent” residents were selected on the basis that they had been permanently resident in the communal establishment for the past six months or, if resident for less than six months, had been in residential care anywhere for at least six months, had no other place of residence at this time, or were likely to remain in residential care for the foreseeable future.5 In smaller establishments (fewer than 80 residents), one in four residents were included in the sample; in larger establishments one in 12 residents were included. Interviews varied: when residents were incapable of answering questions, a member of staff sat in with the subject, or sometimes the administrator was interviewed on behalf of all subjects even though they were competent to answer. In some instances the administrator answered some questions, the subject others.5

The survey collected assessments of disability in locomotion, reaching and stretching, dexterity, seeing, hearing, continence, communication, personal care, behaviour, intellectual functioning, consciousness (fits) and digestion. The Office of Population Censuses and Surveys used the views of carers' organisations, staff, and researchers to scale scores in each of these 13 domains and to develop a composite measure of disability based on the combination of severity scores. The summary measure was calculated as worst + (second worst£0.4) + (third worst£0.3).5 The disability instrument has good interrater reliablility; it is highly correlated with the Barthel index6 but is more comprehensive. 7 8

Our analyses included people aged 65 or over and excluded people for whom “mental handicap” was reported.

General practitioners'workload

For residents of residential and nursing homes, the disability survey collected information about the frequency and typical location of consultations with general practitioners. To compute a measure of workload, we obtained data on the duration of such consultations from the general medical practitioners workload survey 1992-3,9which yielded an estimate of 8.4 minutes for surgery based consultations. For consultations in communal establishments, we used the workload survey's estimate for domestic visits, 13.2 minutes. The mean travelling time for a domestic visit was 12 minutes. We conducted a telephone survey of 38 residential and nursing homes which found that general practitioners see an average of five residents on each visit to a home, equivalent to 2.4 minutes travelling time per resident visit. General practitioners spend time on activities that do not involve direct contact with patients but none the less directly and indirectly contribute to patients' consultations. General practitioners spent a weekly average of 16.9 hours on such activities and conducted 152.1 consultations (equivalent to an additional 6.7 minutes per consultation).9

Where data were missing about whether a person consulted a general practitioner we assumed that no consultation was made. Where information was available on the number of consultations made but not on the typical location, we computed an average of the time for a surgery based consultation (15.1 minutes) and the time for a consultation at the establishment (22.3 minutes), weighted by the proportions of people in all types of communal establishments with completed data, who consulted at the surgery or establishment (5.3% and 94.4% respectively).

For people who had been resident in their communal establishment for less than 12 months, we adjusted the number of consultations to enable comparison with people resident for the entire 12 month study period.

To test the sensitivity of our results to assumptions about general practitioners' workload we conducted analyses for two alternative scenarios: that all general practitioners' visits were to individual residents (31.9 minutes per consultation); and that 90% of consultations took place in a clinic at the establishment, with typically 20 residents seeing the general practitioner en bloc, with shorter consultation times equal to the time for a surgery based consultation, and 10% of visits were to individual residents (average 17.3 minutes per consultation).

Analyses

Consultations with general practitioners differed between accommodation categories, as did characteristics of residents.5 We used multivariate analyses to determine whether variations in utilisation of general practitoners were associated with residents' characteristics or unobserved but systematic differences between accommodation categories (such as internal services or differing links with general practitioners).

The dependent variable was the weekly utilisation per resident of general practitioners' time. There were five groups of explanatory variables:age and sex, ownership of establishment and residential or nursing homes status (based on a survey question which asked whether the home provided “residential services only or residential services with medical, nursing or other professional care”), number of permanent residents, reported underlying long term illnesses, and overall severity of disability.

Statistical methods

The distribution of the dependent variable was non-normal: 15% of residents in the survey did not consult a general practitioner, while the distribution for consulters was skewed with a long tail to the right. To allow for this non-normality we conducted a multistage analysis following the approach of Duan and colleagues to examine the demand for medical care under differing types of insurance 10 11;more recently this has been used to examine the distribution of fundholders' budgets.12

In the first stage a logistic regression was estimated to model the probability that a person consulted a general practitioner. This separated consulters from non-consulters and addressed the first source of non-normality. In the second stage a multiple regression was estimated with the natural logarithm of general practitioner utilisation per resident per week as the dependent variable for those people who consulted. The natural logarithmic transformation reduces the skew in the distribution. The predicted utilisation of general practitioners' time is the product of the predicted probability of consulting (from the first stage) and the predicted utilisation of general practitioners' time (from the second stage) (see Appendix 1 on website).

To allow for differences in residents' characteristics between types of establishment, predictions of general practitioner utilisation were based on utilisation of the entire sample of people in residential and nursing homes, had they been resident in a particular type of accommodation. This was achieved by turning on and off dummy variables (taking the value 0 or 1) as appropriate. We made the predictions more contemporary by substituting mean values for the size of residential and nursing homes from a 1996 survey(32 local authority homes, 17 residential homes, 34 nursing homes).13

The survey estimation routines of the software package STATA were used for the analyses to allow for complex survey design and clustered sampling. For example, there was potentially greater homogeneity of residents within individual homes than in comparison with residents in other homes.

Changes since 1986 in the balance of care

Between 1986 and 1996 the number of elderly people grew, resulting in an increase in the number of disabled people. Contemporaneously, provision of institutional care altered radically. We therefore calculated demographically adjusted estimates of the change in provision. Our first step was to estimate the number of disabled elderly people in Great Britain in 1996 using the age specific prevalence rates for disability (excluding mental handicap) from the disability surveys. The proportions of elderly disabled people in different age groups “permanently” resident in communal establishments in 1986 were again estimated from the disability surveys. We then estimated what number of people would have been resident in the different categories of establishments had they comprised the same proportion of the elderly disabled population in 1996 as they had in 1986. 

Our second step involved estimating the actual numbers of people resident in different categories of communal establishments in 1996 from various official statistics and adjusting them to make them comparable with the disability survey's definition of a permanent resident. The demographically adjusted estimate of the change in provision was equal to the estimates from the first step minus the estimates from the second step (see Appendix 2 on website).

Predicting utilisation by “NHS residents”

We used the sample of people cared for in NHS hospitals in the disability survey to represent people who would previously have been cared for in NHS establishments but who are now cared for outside hospital. We assumed that they would now be cared for in private nursing homes, given the growth in places in this type of accommodation. We predicted general practitioner utilisation by using the equations for private nursing homes. Various assumptions regarding the extent of “downsizing” in NHS provision were examined by taking subsamples of the NHS sample, divided on the basis of the overall severity score for disability under the assumption that the least disabled people were most likely to be cared for outside NHS hospitals.

Patterns of consultation

Compared with their counterparts in households in 1985, disabled elderly people in residential and nursing homes in 1986 consulted general practitioners more often (mean 8.9 v 6.3 times per year) and were less likely to consult at the surgery (5% v 60%).4 These figures are consistent with more recent data for the population (disabled and non-disabled, aged 65 and over). 14 15The finding that residents of voluntary residential homes were more likely to consult general practitioners at their surgery is consistent with evidence that such residents were more likely to make social trips and other excursions outside their home.16

Systematically higher levels of utilisation in nursing homes may not be associated simply with the higher dependency of their residents but may be due to better management of residents' medical care. People who were resident in NHS establishments in 1986 have high predicted levels of utilisation, consistent with previous evidence that elderly people who had been NHS long stay patients had 9.3 consultations per year when discharged to alternative residential settings.17

The presentation of results for alternative scenarios illustrates the sensitivity of our predictions to assumptions about the duration of consultations. It could be argued that more routine consultations at the institution may be shorter. On the other hand, average consultation times may be exceeded by frail elderly people with multiple disorders. Our assumption that where data were missing about consultations the person did not consult a general practitioner is plausible for a well known, commonly used service. Furthermore, we have not included time spent on telephone consultations, which account for 7% of general practitioners' working time.9Taken together, these factors leave our utilisation estimates on the conservative side.

Patterns of provision

We were unable to include supply-side effects, and our results implicitly assume that general practitioners met the extra demand and that the pattern of provision has not changed since 1986. But primary care has changed —for example, the number of practice nurses increased rapidly18and now accounts for one in eight consultations,14and there has been a downward trend in the proportion of consultations taking place outside the surgery.15 However, the 1986 survey describes people in NHS establishments before the major shift in provision occurred and so represents the kind of people previously cared for in hospital but now in residential and nursing homes. Recent research confirms the increasing dependency levels of residents of residential and nursing homes.19

The explanatory variables for disability and morbidity were associated with the use of general practitioners. Nevertheless, some differences in morbidity were not captured by these variables. In this context, the high predicted utilisation for voluntary nursing homes needs to be viewed with caution, especially in view of the relatively small sample size.

Policy implications

Two policy implications of these results can be identified. At a macro level, as beds closed, financial transfers from hospital budgets to other budgets varied enormously in size and orientation across the country, but we are unaware of any transfers to primary care. Changing provision of long term care for elderly people has accounted for a considerable proportion of the increased number of general practitioners 20 21while deinstitutionalisation of psychiatric patients and demographic change have also increased workload. The change in demand will be pronounced for practices in areas with an older population, but territorial equity will be hard to achieve while general practitioners remain independent contractors.22

At a micro level, primary care groups23will receive resources from (merged, larger) health authorities but general practitioners will continue to be independent contractors (although the introduction of the salaried scheme is significant in this context). Defining the appropriate activities of general practitioners as independent contractors as opposed to commissioners and providers of community health services will be critical to the quality of care for elderly people in institutions and, more generally, to the success of primary care groups. Although family health services authorities and health authorities have already been merged, the flexibility of the new arrangements has been questioned.24Refined funding formulas to allocate funds to primary care groups and to general practitioners as independent contractors may be required —but however refined the formula, the effect will be to shift financial risk to the groups and general practitioners. In this context, care outside hospital of ever more chronically ill people, imposing large demands on general practitioners, may lead to the undertreatment or disenrolment of such expensive patients. 12 25Indeed, some general practitioners were unwilling to enrol former long stay patients on their lists (even before the introduction of fundholding).17 In this context, some homes are paying extra sums to general practitioners for the care of their residents. These charges, passed on in extra fees to local authorities and residents, blur the health and social care boundary and introduce means tested charging for health care.

In the United States some evidence suggests that health maintenance organisations provide poor coverage and treatment for chronically ill people. 26 27Contracting arrangements that share the risks between health authorities and primary care groups may be appropriate.28This could involve the health authority and the group sharing the costs above the target amount set by the formula (for example, 50:50), or an extension of the current stop-loss arrangements where an expenditure limit of £6000 per patient applies to fundholders. The two arrangements could be combined. Another possibility would be to separate (carve out) arrangements for chronically ill groups.28