Interpersonal continuity of care: a cross-sectional survey of primary care patients' preferences and their experiences

Patients

The study was conducted in Leicestershire and West London, locations with ethnically and socioeconomically diverse populations, varied local services, and rural as well as inner city settings. Patients were identified through the lists of 22 practices (nine in London, 13 in Leicestershire) selected to ensure diversity from those expressing interest in participation following a letter of invitation to all practices in Leicestershire and Kensington and Chelsea. Practices were asked to draw a random sample of 50–250 people aged 18–80 years according to practice size, and stratified into four age groups (18–29, 30–54, 55–74, and >75 years), since healthcare needs tend to vary according to age. Practices posted the questionnaires with a covering letter signed by a doctor, with reminders sent to patients after 2 and 4 weeks. Questionnaires were also issued to consecutive patients of the Leicestershire walk-in centre. A sample size was calculated to ensure, within each of four age groups, 95%confidence intervals for prevalence estimates of +/−5%. This required 400 responders from each group.

The following information was collected about participating practices by interviewing practice managers: number of registered patients; appointment system; proportion of consultations delivered by nurses; practice culture relating to information sharing and teamwork;13 and practices' attitude towards continuity of care. The practice Index of Material Deprivation (IMD) 2004 score was used to indicate socioeconomic deprivation, a higher score indicating greater deprivation (mean score for areas in England 21.7).14

The questionnaire

The 31-item questionnaire included questions on age; sex; ethnic group;15 employment; education; carer status; time registered with the practice; services used in the past year (GP, nurse, out-of-hours, accident and emergency, walk-in centre, NHS Direct and pharmacist); their most recent consultation (who with; the service consulted, reason either new, routine/review of long-term condition, or other such as health promotion); social support and social integration (feeling part of the area lived in and contact with friends or family in the last 2 weeks16^,17); presence of long-term illness; and EuroQol (EQ–5D, low score indicates worse health).18 Questions were asked about nine aspects of care with respect to responders' most recent consultation:

• three on elements of interpersonal continuity —choosing a particular person, choosing someone known and trusted, and choosing someone who knows personally the patient and medical condition;

• one on being able to book an appointment in advance;

• four on the person consulted — type of professional (for example choosing a nurse or a doctor), someone of the same sex, someone of the same ethnic group or culture, someone who would take time to listen; and

• one on informational continuity — someone with paper or computer notes containing the full medical history of the patient.

For each question, the responder was asked to indicate importance (four-point option format, extremely important to not important), then whether they had actually received that aspect of care (yes/no response). They were also asked when they had wanted their consultation and when it had actually taken place.

Analysis

The categories for importance of the nine aspects of primary care were collapsed into two groups (extremely important and important, and slightly important and not important), and combined with the second element to produce three possible responses: the aspect of care was not important; important and experienced; or important but not experienced.

Bayesian Markov chain Monte Carlo methods were used because of problems estimating models using traditional logistic regression with sparse data in some cells. Models containing random effects for practices were fitted to allow for clustering. Following initial univariable analysis, potential explanatory variables were included in two-level multilevel logistic models (level one the patient, level two the practice), which were fitted to the data for each dependent variable separately using MLWin 2.01. Seventeen binary variables were investigated, nine relating to whether the responder wanted that aspect of care, and eight to whether the responder who wanted the aspect of care did or did not receive it. Each independent variable was entered into a univariable model as a predictor for each dependent variable. Only significant predictors from these univariable models (using parameter estimates together with 95% Bayesian credible intervals to determine significance)19 were then entered into further models, in the process of implementing forward selection for each dependent variable separately. The final model in each case contained only effects significant at the 5% level. Given the number of hypotheses being tested the probability of type I errors is likely to be high, but since the study was exploratory rather than strictly hypothesis-testing no adjustment was made for the large number of tests. Modelling was undertaken using Monte Carlo methods, although starting values were derived using Marginal Quasi-likelihood methods, after which the Monte Carlo method was implemented to derive the estimates, their standard errors and Bayesian 95% credible intervals. The Bayesian approach constructs a credible interval, centred close to the sample mean, but affected by prior beliefs concerning the mean. There is a 95% probability that this interval contains the true mean. Monte Carlo methods allow Bayesian models to be fitted with prior parameter distributions. By default MLwiN sets vague priors, and these were the priors (γ, α = 0.001, β = 0.001) used in the current set of analyses.

Each model was run for at least 50 000 iterations. For each dependent variable intercepts and slopes were allowed to vary randomly. After final models had been chosen, the results were checked by refitting each model twice, with no random effects and with random intercepts only. The Bayesian deviance information criterion was used to decide which of the resulting models was best, the model having the smallest criterion being considered superior. The criterion takes into account how well the model fits the data and also the complexity (parsimony) of the model.

Summary of main findings

Most patients experienced the aspects of care they regarded as important, and this may be interpreted as a success for general practice, but nevertheless obtaining interpersonal continuity was difficult for some patients who wanted it. Seeing someone known and trusted was important to 62.6% of responders, seeing a particular person important to 65.2% responders and seeing someone who knows the patient and medical condition personally was important to 75.3% responders. Of those who regarded these features of interpersonal continuity as important, 13.7, 21.3 and 17.0%, respectively, failed to experience them. The majority of patients also wanted to consult someone perceived as taking time to listen and with information about their clinical history (informational continuity), and most (but not all) experienced these attributes of care at their most recent consultation.

Groups that were more likely to fail to get what they wanted were people in work (seeing someone with information, someone with time to listen, someone known and trusted, someone who knows the patient and condition), being non-white (someone with time to listen) and being socially isolated (someone of the same sex). Thus, people in work and people who are not in work for any reason other than retirement have more difficulty experiencing informational and interpersonal continuity than people who are retired, while nonwhite ethnic groups and people who are socially isolated have difficulty negotiating other desired aspects of care.

Strengths and limitations of the study

The response rate was similar to the response rate of 47% for the 2005 national patient survey, in which rates varied from 23 to 61% between primary care trusts.20 A wide range of patients, including those from different ethnic and socioeconomically disadvantaged groups, were incorporated into the study and in future surveys researchers may have to choose between achieving high response rates from relatively homogenous advantaged populations or lower rates from more diverse populations. In the 2005 national survey, 5% of responders were nonwhite, but in this study the proportion was 10%. Particular caution is needed, however, in interpreting the findings in relation to the younger age group, who are less likely to have chronic illness and likely to place higher priority on access than continuity. This survey did not involve a national random sample, and the direct extrapolation of the findings to the national population would be inappropriate.

Although this study relies on self-reports about one consultation in the context of continuing relationships, it was informed by the qualitative and longitudinal studies that had preceded it, and which enabled the utilisation of a detailed appreciation of the issues important to patients.8 Furthermore, continuity was classified into interpersonal and informational, and contrasted with other key attributes of primary health care. However, information about practices was reliant on the perceptions of one member of the practice team.

Comparison with existing literature

These findings support the findings of the qualitative10 and longitudinal8 studies in highlighting the role of the patient in contributing to the level of continuity they obtain. Patients with new, minor problems tend to prioritise speedy access before interpersonal continuity, and often also before informational continuity. Patients with long-term or more complex problems place greater weight on informational and interpersonal continuity, the other attributes of the service giving way to the preference for an established relationship.11^,12^,21

In general, retired people had least difficulty in negotiating their wants, perhaps because they have fewest personal time constraints. People who are not in work but not retired, or in a non-white ethnic group, or who have a degree of social isolation tend to have greater difficulty in obtaining what they want. The way in which practices and local health services operate appears to discriminate against these disadvantaged groups. It could be that they have too many conflicting priorities or are less skilled at negotiating their preferred appointments.

Implications for research and clinical practice

While it is good to find that so many responders achieved the interpersonal continuity they wanted, there is still room for improvement for others. Service providers must attend to the needs of disadvantaged groups and take steps to help them obtain the primary care they prefer, particularly from someone they know and trust. Such steps should involve minimising the complexity of service design and operating flexible appointment systems that include the option of booking appointments in advance, as proposed in the recent government white paper's.3 Better training for receptionists in sensitivity to the needs of people who have difficulty in negotiating their preferences for interpersonal continuity is another way forward.