|Home | About | Journals | Submit | Contact Us | Français|
Although continuity of care is a widely accepted core principle of primary care, the evidence about its benefits is still weak.
To investigate whether continuity of care in general practice is associated with better survival in older people.
Data were derived from the Longitudinal Aging Study Amsterdam, an ongoing cohort study in older people in the Netherlands. The study sample consisted of 1712 older adults aged ≥60 years, with 3-year follow-up cycles up to 17 years (1992–2009), and mortality follow-up until 2013.
Continuity of care was defined as the duration of the ongoing therapeutic relationship between patient and GP. The Herfindahl–Hirschman Index was used to calculate the continuity of care (COC). A COC index value of 1 represented maximum continuity. COC index values <1 were divided into tertiles, with a fourth category for participants with maximum COC. Cox regression analysis was used to investigate the association between COC and survival time.
Seven hundred and forty-two participants (43.3%) reported a maximum COC. Among the 759 participants surviving 17 years, 251 (33.1%) still had the same GP. The lowest COC category (index >0–0.500) showed significantly greater mortality than those in the maximum COC category (hazard ratio (HR) = 1.20, 95% CI = 1.01 to 1.42). There were no confounders that affected this HR.
This study demonstrates that low continuity of care in general practice is associated with a higher risk of mortality, strengthening the case for encouragement of continuity of care.
Continuity of care denotes the connected and coherent care that is consistent with the health needs and personal circumstances of a patient.1,2 It is considered important to ensure effective and efficient health care and is believed to be essential for high-quality patient care.3
Three major types of continuity of care are commonly distinguished, namely management, informational, and relational continuity.1,2,4,5 Management continuity refers to multidisciplinary and institutional coordination and coherency of the delivery of complementary healthcare services to achieve health goals.4,6 Informational continuity concerns the availability of previous information among different healthcare providers that is appropriate to the current circumstances of the patient.2,4,7 Relational continuity, which is most valued in general practice, refers to the ongoing therapeutic relationship between the patient and one or more healthcare providers that bridges episodes of care.2,4,5
Nowadays, continuity of care is a widely-accepted core principle of primary care.6,7 The assumed benefits of continuity of care include a better patient–provider relationship, increased patient satisfaction, improved uptake of preventive care, enhanced adherence to treatment, more accessible health care, and reduced healthcare use and costs.3,6–17 Especially vulnerable patients, such as older patients, are considered to benefit from continuity of care, as they are likely to have multiple chronic conditions.6,18 According to Haggerty et al,2 improving continuity of care has become a research priority, as patients increasingly receive care from multiple professionals and organisations.
Despite the assumed benefits of continuity of care, most studies are based on patients’ experience and have a limited sample size due to the burden of data collection.7 According to Wolinsky et al,19 mortality may be the most appropriate criterion to measure the effect of continuity of care, especially in older people. To date, only three studies have investigated the relationship between continuity of care in primary care and mortality. Leleu and Minvielle7 performed an observational study based on reimbursement claims from the French national health insurance system and found higher continuity of care to be associated with a reduced likelihood of death (adjusted hazard ratio (HR) = 0.96 [95% confidence interval (CI) = 0.95 to 0.96). Wolinsky et al used data from the Survey on Assets and Health Dynamics among the Oldest Old (AHEAD) in the US and concluded continuity of care to be associated with substantial reduction in long-term mortality (adjusted HR = 0.84 (95% CI = 0.77 to 0.91)).19 In a retrospective cohort study among older patients with diabetes in Canada, the higher-continuity group had lower rates of death than the lower-continuity group (8.6% versus 18.5%).20 Although these three studies found a beneficial effect of continuity of care on mortality, the benefits of continuity of care may depend on the national healthcare context.
Although continuity of care is a widely-accepted core principle of primary care, Evidence for its benefits is still weak. The present study demonstrates that low continuity of care in general practice is associated with a higher risk of mortality, strengthening the case for encouragement of continuity of care. This study adds to previous research by providing follow-up to 17 years, face-to-face interviews, and broad data coverage.
Using data from the Longitudinal Aging Study Amsterdam (LASA), this study investigated whether continuity of care in general practice is associated with mortality in Dutch older people. The study was carried out in the Dutch healthcare system, which is characterised by the GP acting as gatekeeper and patients covered by general health insurance.
Mortality was used as the outcome, because death is considered the ultimate criterion to measure the effect of continuity of care.19
This study adds to previous research by providing follow-up to 17 years, face-to-face interviews, and broad data coverage. It was hypothesised that discontinuity of general practice care is associated with a higher risk of mortality.
The LASA is an ongoing cohort study on physical, emotional, cognitive, and social functioning in older people in the Netherlands. The LASA cohort (n = 3107) was recruited in 1992 from a random sample of older men and women aged 55–85 years, in the west, north east, and south of the Netherlands. The sample was stratified by age, sex, degree of urbanicity, and expected 5-year mortality. The LASA cohort is representative of the older Dutch population with respect to geographic region and degree of urbanicity. Sampling, data collection, and non-response are described elsewhere.21,22 Since 1992, longitudinal data have been collected every 3 years. Measurements are performed by trained interviewers, who visit responders at home.
For the present study the first six cycles of data collection were used (1992–1993, 1995–1996, 1998–1999, 2001–2002, 2005–2006, and 2008–2009). Responders were included (n = 1712) if they participated in at least three subsequent data collection cycles (starting in 1992–1993) and if there was complete data on their GPs. The time period across which continuity of GP care was determined varied from 7 years (1992–1999) up to 17 years (1992–2009).
Informed consent was obtained from all participants.
Information about mortality was obtained through linkage with registers of the municipalities in which the responders were living (most recent probing date: 1 November 2013).23 Mortality follow-up was incomplete for four cases. Survival time was computed as the date of death or the most recent probing date minus the interview date of the last data collection cycle for which the GP was known.
Continuity of care (COC) was defined as the duration of the ongoing therapeutic relationship between patient and GP, that is, relational COC. This type of COC is most valued in primary and mental health care.1 To calculate the COC, the Herfindahl–Hirschman Index (HHI) was applied:24,25
The HHI is an economic measure of concentration, defined as the sum of squares of market shares, whereas market shares are expressed as fractions.26 In the present study, is the ‘market share’ of GPi and N is the number of different GPs for a specific patient. The number of data collection cycles in which a participant provided the name of a specific GP were counted. Fractions were allowed because participants sometimes reported two different GPs per data collection cycle (both GPs were then counted as ½). This count was subsequently divided by the total number of data collection cycles in which the participant provided the name of a least one GP. These ratios were squared and then summed, resulting in the participant’s COC index. The COC index ranges from 1/N (minimal continuity, a different GP at each data collection cycle) to 1 (maximal continuity; the same single GP in all data collection cycles). For example, for a participant reporting GP A (cycle 1 and 2), GP B (cycle 3, 4, 5, and 6), and GP C (cycle 3), the COC index is calculated as: contribution GP A + contribution GP B + contribution GP C = (2/6)^2 + (3.5/6)^2 + (0.5/6)^2 = 0.458.
Based on previous research, the covariates included were age, sex, sociodemographic characteristics, smoking, alcohol use, morbidity, functional limitations, depression, cognition, and personality characteristics.7,12,19,27–38 Covariates were measured at the last cycle for which GP data were available.
Sociodemographic characteristics included housing (independent or care institution), level of education, level of urbanicity, and partner status. Level of education was divided into low (elementary school or less), middle (lower vocational, general intermediate, intermediate vocational, or general secondary school), and high (higher vocational education, college, or university).39 Level of urbanicity was assessed using the number of addresses per square kilometre, distinguishing five categories from low (<500) to very high (>2500).40 Partner status was divided into three categories: no partner, co-residing partner, or partner outside of household. Smoking status was classified as nonsmoker, former smoker, or current smoker..30 Alcohol use was assessed by using the Garretsen indicator of alcohol use, distinguishing three categories: no, light, and moderate to very excessive use of alcohol.41
Morbidity was assessed by self-report of seven major chronic diseases: chronic pulmonary disease, cardiac disease, peripheral arterial disease, diabetes mellitus, stroke, osteoarthritis or rheumatoid arthritis, and cancer. These self-reported answers have been found to correspond well with information from GPs across the full study period.42
Functional limitations were assessed by six self-reported questions about experienced difficulty in doing daily activities, counting the number of items ‘with some difficulty’ or worse (range 0–6; internal reliability 0.85).43 These activities included walking up and down stairs, walking outside the house, use of transportation, dressing oneself, sitting down and rising from a chair, and cutting own toenails.
Cognition was assessed by means of the mini-mental state examination (MMSE; range 0–30; cutoff for cognitive impairment ≤24).46
Personality characteristics were assessed by sense of mastery, self-efficacy, and self-esteem. Sense of mastery is defined as the extent to which one views one’s life as within one’s control as opposed to being ruled by chance or other people. It was measured by the seven-item Pearlin Mastery Scale (range 7–35).47 Self-efficacy is defined as the belief of a person in their ability to organise and execute certain behaviours that are necessary to produce given attainments. It was measured by a 12-item version of the General Self-Efficacy Scale (GSES- 12; range 12–60).48 Self-esteem reflects a person’s overall evaluation or appraisal of their own worth. It was measured by an adapted four-item version of the Rosenberg Self-Esteem Scale (range 4–20).36,49
The association between COC and survival time was investigated using Cox regression analysis. In preliminary analyses there was a non-linear association between the COC index and mortality. Therefore, it was divided into four categories. The first category included the COC index value of 1 representing maximum continuity. COC index values <1 were divided into tertiles. Bivariate comparisons were performed to examine the associations of the main outcome mortality with all covariates.
Confounding was investigated by manually introducing all covariates into the basic model. A covariate was considered a confounder if the coefficient of COC changed by more than 10%. Effect modification was investigated by calculating interaction terms of each covariate with COC. Dummies were used to investigate categorical covariates. For COC, the group with maximum COC (index = 1) was used as reference group. Responders participated in three, four, five, or six data collection cycles, corresponding with different periods of data collection. Survival time was calculated starting from the interview date of last data collection. Although this approach made optimal use of available data, responders participating in more data collection cycles may have been younger and healthier, but at the same time their survival time was calculated from a later interview date. In addition, participants who died relatively early may have had a limited number of data collection cycles. To account for this, the patient’s final wave of data collection was used as a stratification variable and allowed baseline hazards to differ between these strata.
The level of statistical significance was set at P<0.05. Data were analysed using SPSS (version 22).
Seven hundred and forty-two participants (43.3%) reported a maximum COC, indicating that almost half of older adults had the same GP for at least 6 years. Frequencies for low, moderate, and high COC were 370 (21.6%), 290 (16.9%), and 310 (18.1%), respectively, after the third data collection cycle. Among the 759 participants surviving 17 years, 251 (33.1%) still had the same GP (Table 1). The mean and median of the COC index were 0.74 and 0.72.
The COC index did not show differences across characteristics as measured during the last cycle at which GP data were available, except for level of urbanicity (P<0.006; Table 2). In less urban areas COC tended to be higher than in highly urban areas.
In bivariate analysis, the covariates male sex, higher age, institutional care, no partner, higher level of urbanicity, current smoking, moderate to very excessive alcohol use, three or more chronic diseases, functional limitations, impaired cognition, depressive symptoms, lower mastery, and lower self-efficacy were associated with mortality (Table 3). No confounders were found that changed the hazard ratios of the main determinant by more than 10%.
As participants with six data collection cycles (that is, period of data collection and shortest period since the last GP registration) had a better survival compared with other participants (Appendix 1), they were stratified by the number of data collection cycles (six versus three, four, or five cycles). In the final model, participants in the lowest COC category showed significantly greater mortality than those in the maximum COC category (HR = 1.20, 95% CI = 1.01 to 1.42; Table 4).
The number of participants in this study decreased from 1712 (100%) in 1992–1993 to 759 (44.3%) in 2008–2009. Attrition in LASA can be attributed for the largest part to mortality, and to a lesser extent to refusal, frailty, or the inability to communicate.22
This study investigated the association between continuity of care in general practice and mortality in older people during 17 years of follow-up (1992 to 2009). Continuity of care (COC) was defined as the duration of the ongoing therapeutic relationship between patient and GP, and was calculated using the Herfindahl–Hirschman Index. The lowest COC category (index >0–0.500) showed a 20% increased likelihood of mortality compared with the category representing maximum COC. According to the study hypothesis, discontinuity of general practice care was found to be associated with a higher risk of mortality.
Of the few studies that have investigated the relationship between continuity of care in general practice and mortality,7,19 the present study is the first to provide a followup period up to 17 years. Other strengths of the study are the representative sample, the continuity of care calculation that allows fractions (that is, having more than one GP at the same time), and multiple follow-up periods per patient. Another strength is the broad data coverage of the LASA database, which enabled investigation of important associations with mortality and continuity of care, and to adjust for relevant confounders. The study’s confirmation of well-known associations between mortality and each of the variables sex, age, partner status, smoking, morbidity, functional limitations, cognition, depression, and mastery illustrates the robustness of the data.
For the interpretation of the results, some limitations need to be considered. First, the study focused on relational continuity of care and defined continuity of care as the duration of the ongoing therapeutic relationship between a patient and their GP(s),1,5 using a calculation that takes into account the number of GPs, change(s) of GP(s), and the length of the respective therapeutic relationships. However, there are many other ways to define and measure continuity of care, such as concentration (the proportion of consultations with one specific provider) and sequence (whether each consultation was with the same provider as the previous consultation).1,5,11,50
A limitation of the study’s continuity of care measure is that it does not take into account how often, or how seldom, a patient may see their provider. For example, a patient for whom a long-term therapeutic relationship was observed may never seek care from that provider and hence will have very little interaction with that provider. Insight into the frequency with which care was sought (and with whom) would have enabled alternative continuity of care measures to be calculated, therefore enriching the data and their interpretation. The LASA database unfortunately does not provide data on consultation level.
Second, a change of GP may be due to different — intentional and unintentional — reasons, including preferring another health provider (patient), retirement (GP), moving (patient or GP), or conflict (patient and GP). As different reasons for change may be associated with a different quality of the patient–provider relationship, one may suggest that adding the covariate ‘reason for change of GP’ could improve the final model. However, given the absence of supportive evidence, it is unlikely that the covariate ‘reason for change of GP’ is a confounder (that is, that it correlates with both continuity of care and mortality) and therefore unlikely that adding this covariate will significantly change the final model.
Three previous studies found a beneficial effect of continuity of care on mortality in a primary care population. Leleu and Minvielle performed an observational study based on reimbursement claims from the French national health insurance database for salaried workers between 2007 and 2010.7 They found higher continuity of care in primary care to be associated with a reduced likelihood of death (adjusted HR = 0.96 [95% CI = 0.95 to 0.96]). Wolinsky et al used data from the Survey on Assets and Health Dynamics among the Oldest Old (AHEAD) with up to 12 years of follow-up between 1991 and 2005.19 They concluded that continuity of care with a primary care physician was associated with substantial reduction in long-term mortality (adjusted HR = 0.84 [95% CI = 0.77 to 0.91]). In a retrospective cohort study of patients with diabetes aged ≥65 years, the higher-continuity group had lower rates of death (8.6% versus 18.5%) than the lower-continuity group.20
As the present study showed, patients with relatively short-term therapeutic relationships with several GPs had a shorter survival than patients with the same single GP during the full follow-up period. These results add to the studies of Leleu and Minvielle,7 Wolinsky et al,19 and Worrall and Knight:20 high continuity of care is related to lower mortality, and this relationship appears to exist across different healthcare contexts and for different measures of continuity of care. Comparing the group characteristics of low, moderate, high, and maximum continuity of care does not reveal another explanation, as only few characteristics were found to be associated with continuity of care (Table 2).
Although continuity of care is a widely-accepted core principle of primary care, the evidence about its benefits is still weak, especially when it comes to health outcomes. The present study demonstrates that low continuity of care in general practice is associated with a higher risk of mortality, strengthening the case for encouragement of continuity of care. The theoretical framework of continuity of care, however, is complicated and many components are involved — at patient, doctor, and system level. Further research should acknowledge this complexity and measure the involved components separately,50 to differentiate the good and adverse effects of continuity of care.
The Longitudinal Aging Study Amsterdam is largely funded by grants from the Ministry of Health, Welfare, and Sport of the Netherlands. The sponsors did not participate in the study design, data collection, analysis, interpretation, or the preparation or submission of this report. The first author (Otto R Maarsingh) was supported by an individual grant of the Stichting Beroepsopleiding Huisartsen (SBOH) in the Netherlands.
The medical ethics committee of the VU University Medical Center approved the study (reference nos 92/138 and 2002/141).
Freely submitted; externally peer reviewed.
The authors have declared no competing interests.
Contribute and read comments about this article: bjgp.org/letters