Background

Canadian provincial and federal health commissions have concluded that a strong primary healthcare foundation is the key to a sustainable health system [1–6]. Ecologic studies suggest that regions with robust support in primary healthcare have better health indicators, such as longer life expectancy, lower all-cause mortality, better health equity [7–13] and show better intermediate outcomes of care.

For the purposes of this research we use "primary healthcare" in its narrow sense to refer to primary medical care, provided in organizational models composed minimally of family physicians or generalists, who may or may not be working with other health and social services professionals.

The goal of primary healthcare is to optimize health and functional health through activities of timely diagnosis and treatment, clinical disease prevention, health promotion and support during rehabilitation and palliative care. Through effective care, clinicians can help patients adopt more positive health behaviours [14, 15], avoid morbidity [16–18] and improve functional health even in the face of prevailing disease [15]. As patient-centeredness is the core value of care delivery among primary healthcare professionals [19, 20], this research includes dimensions of primary healthcare identified as being "person-oriented": [21] accumulated knowledge of the person (relational continuity), interpersonal communication, shared decision-making, and respectfulness.

While quality and effectiveness of healthcare is usually measured by provider's compliance with established norms for defined conditions, this research focuses on quality as it relates to the person-centeredness of clinical processes and approaches and their impact on functional health and health behaviour [22–25].

Functional health is the extent to which an individual perceives that physical or mental health limits his/her capacity to carry out daily activities and social roles [26, 27]. Functional health declines with increasing chronic illness burden [28–30]. The repeated Canadian Community Health Surveys (CCHS) demonstrate that the population prevalence of chronic illnesses is increasing [31, 32] but provide little insight about how the increase of illness burden in individuals is dynamically related to functional health. Functional health can improve through self-management and self-efficacy to change negative health behaviours and adopt positive ones [30, 33].

Although health behaviour change is influenced by multiple factors, there is relatively strong evidence that physicians' recommendations and reinforcement have a strong influence on such changes, e.g. smoking [34, 35], alcohol consumption [36–38], regular exercise [39–42] and healthy weight. In this research we will study both health behaviour status and intention to change.

There are groups of patients who are at high risk of health deterioration and may be particularly vulnerable to problems in the organization of healthcare such as the multi-morbid and the poor and consequently, most likely to benefit from patient-centred care. Our rough estimates suggest that the population prevalence of multi-morbidity increases by 1% per year of age [43–47]. Multi-morbidity is a major issue for primary healthcare providers [43] as research shows clear associations between multi-morbidity and the experience of unfavourable outcomes. We also found a clear association between illness burden and functional health and psychological distress [48–52].

Poverty is a state of material and/or social deprivation that limits the capacity to mobilize resources to achieve well-being [53, 54]. The way care is delivered has a large impact on the effectiveness of care that the poor receive. Although ecologic studies suggest that primary healthcare can improve health inequity in the population [7, 9–13], the demonstration for what impact this may have can only be made through the longitudinal follow-up of individuals.

The underlying premise of this research is that patient-centred and effective primary healthcare can maximize functional health, in general, and particularly in vulnerable groups, such as those with a high burden of chronic illnesses and the poor. We will longitudinally follow the health and healthcare experience of individuals: 1) to describe changes in functional health, chronic illness burden and health behaviours; 2) to examine the impact of patient-centred and effective primary healthcare on functional health and other outcomes of interest (health behaviours, chronic illness burden, health service utilization); and, 3) to explore the relationships between intermediate outcomes and individual characteristics, and functional health.

Methods/Design

The proposed study is a cohort of 2000 adults aged 25-to-75 years followed for 4 years. The target population is community-dwelling adults undifferentiated by disease, who would seek primary healthcare locally, do not suffer from major cognitive impairment, and are able to respond to written and oral questions in English or French. Participants will be randomly selected within the geographic boundaries of four local healthcare networks in metropolitan, rural and remote urban agglomerations of Québec. At recruitment (T0), cohort participants will report on socio-demographic information, functional health and healthcare use. Two weeks (T1), 12 months (T2) and 24 months (T3) after recruitment, they will complete a self-administered questionnaire on their current health, health behaviours and primary healthcare experience in the previous year. Use of medical services will be confirmed through the review of administrative databases.

Participants will be recruited through a telephone survey with a two-stage sampling design. Following first contact, staff will select the adult in the household with the most recent birthday [55]. Participant contact information will be sent to the research team (independently of data), who will then mail a "welcome package" containing a consent form, questionnaires and a postage-paid return envelope. Phone contact will follow to review the consent form and respond to questions.

Follow-up and cohort maintenance

The principal threat to the internal validity of a cohort design is the differential loss to follow-up. We will optimize cohort maintenance and subject retention by using newsletters and greeting cards but do expect some attrition between recruitment and the return of the T1 questionnaire, and over time [55]. To have 2000 subjects at T3, we will initially over-recruit by 20% (2400 at T1).

Data collection

Patient self-report questionnaires

T0

Demographic data and information on functional health and use of health services over the previous year collected at T0 will reduce the later response burden and provide valuable information on patients lost to follow-up, namely age, gender, language, education, perceived income adequacy, usual source of primary healthcare and the strength of affiliation, and overall assessment of health.

T1, T2, and T3

The self administered questionnaire containing approximately 160 questions will be available on paper (mailed) or internet. Since online responding allows for immediate data capture and built-in quality checks, we will strongly encourage this modality [56]. Respondents with chronic diseases will respond to an additional set of 32 questions. Overall, it takes approximately 50-70 minutes to complete (general vs. chronic disease) [57]. We used validated subscales where possible, as outlined in Table 1, and described in detail below for key components.

Table 1 Operational definitions, measurement instrument, and available metric properties of primary and secondary outcomes of interest
Full size table

We will apply the Dillman method [58] to maximize response to questionnaires at T1, T2 and T3: a personalized reminder/thank you note (postal or e-mail) at 2 weeks, followed by a re-mailing of the questionnaire to non-responders at 4 weeks, followed by a reminder to continued non-responders at 6 weeks and a phone call at 8 weeks. Compensation will be mailed with the questionnaire to enhance response [59]. Subjects will be considered lost-to-follow-up after eight weeks of non-response or explicit refusal to continue to participate.

Administrative medical services

We will use administrative medical services data from the Quebec healthcare insurance agency (RAMQ) to identify emergency room visits, hospitalizations and specialist visits, as secondary outcomes of interest.

Outcomes of interest

The advantage of a prospective cohort design is the capacity to examine multiple outcomes of interest; some of which may be conceived as independent or mediating variables for other outcomes. Due to space limitations, we only provide operational definitions for functional health, our main dependent variable and two other intermediate outcomes: chronic illness burden and health behaviours. Figure 1 displays the conceptual model of the study. Table 1 provides an overview of the operational definition and available metrics of all outcomes measured when it applies.

Figure 1
figure 1

Conceptual Model.

Full size image

Functional health

The main dependent variable in this project is functional health status measured with the second version of the Short-Form-12 survey (SF-12v2) [60, 61]. It distinguishes between degrees of good health and poor health [62] and is sensitive to mild changes in illness burden [63]. It will allow us to examine the physical component, the mental component and overall assessment of health status separately. Functional health status is elicited by questions on physical health (physical functioning, role limited by physical capacity, bodily pain, overall health) and mental health (emotional health, vitality, social functioning, role limited by emotional state) in the last 4 weeks.

Chronic illness burden

We will measure illness burden using the validated Disease Burden Morbidity Assessment as this tool provides us with more sensitive and specific data than chart reviews [64]. For each of 22 physical and mental conditions diagnosed by a health professional, the person reports the extent to which the illness interferes with daily activities. Changes in score reflect both number of diseases and their perceived impact on daily living; consequently, both increases and decreases can occur over time.

Health behaviours

We will measure the presence and intensity of health behaviours (vegetable consumption, smoking, alcohol consumption, healthy weight and physical activity) using validated sub-scales from the Behaviour Risk Factor Surveillance System Questionnaire [65], from the Enquête Saguenay-Lac-St-Jean 2007 [66] and other regional surveys, and the CCHS questionnaire. We propose a summary score of health behaviour status, with negative scores for negative health behaviours and positive scores for positive health behaviours, ranging from -8 to 6. In addition, we will measure self-reported intent to engage in or adopt each healthy behaviour, using a single-item five-point response scale [67, 68] that maps validly to the stage-of-change model [69] and has been linked to both functional health and future behaviour [70–72]. The intention scores used in this model predict long-term behaviours and are less labile than actual behaviours.

Predictor variables

The main independent variable of interest is the patient's reported experience on the different dimensions of patient-centred and effective primary healthcare received from the regular provider over the previous 12 months, elicited at T1, T2, and T3. In addition, we are interested in the confounding and modifying effects of individual characteristics, especially multi-morbidity and poverty, but also other characteristics such as age effects and social support. The operational definition, sub-scale and available metric properties of these variables are outlined in Table 2.

Table 2 Operational definitions, measurement instrument, and available metric properties of predictor variables
Full size table

Patient-centred care

Our principal measure is the Perception of Patient-Centred Care [73], adapted for usual care rather than for a single visit. We will further supplement this by exploring related dimensions such as relational continuity [74], interpersonal communication [74], shared decision-making and respectfulness [75]. All subscales refer to usual care. They are principally informative and accurate in identifying those who have a negative experience.

Effective care

Dimensions of effective care are patient perceptions of accessibility, coordination, prevention and health promotion, chronic illness care and patient safety, over the previous 12 months. We will measure accessibility through experienced timeliness of first contact care for urgent (but not emergency) problems [76], organizational flexibility for accommodating urgent care [77], and overall organizational accessibility [74]. Coordination is measured only in those who have seen more than one provider and measures the extent to which care is experienced as connected and coherent [78]. Measures of prevention and health promotion are measured by patient recall of the provider conducting specific clinical preventive activities and addressing the life-style habits we are measuring in our health behaviour score. The chronic illness care scale measures the extent to which elements from the Chronic Care Model [79] have been implemented by all the providers [80]. Finally, patient safety is measured by using indicators of medication errors and the receipt of risk-reduction, clinical and educational manoeuvres.

Multi-morbidity

will be inferred from the validated Disease Burden Morbidity Assessment [64]. Based on our conception of multi-morbidity, we propose an operational cut-off score at >10, corresponding to several diseases with minimal impact on daily living or at least two with major impact. However, a secondary objective of our analysis is to identify the threshold which is most sensitive to declining functional health, reflecting the current stage of development of multi-morbidity.

Poverty

will be based on the Statistics Canada low income cut-off for households, adjusted for household composition [81]. This corresponds to family incomes where the expected expenditure on food, shelter and clothing is 20 percentage points higher than for the average family. We will also generate a composite score of economic vulnerability using highest educational achievement, employment status, housing, per capita household income and perceived income adequacy.

Analysis

The unit of analysis is the individual patient followed over the study period. We will conduct cross-sectional analysis to evaluate the comparability of our study sample with CCHS samples for Quebec and Canada. We will also confirm previously-described relationships between individual characteristics and chronic illness burden, health behaviours and functional health, as well as cross-sectional associations with healthcare.

To estimate the degree of changes in health and health functioning over time (objective 1), we will estimate annual increase in chronic illness burden, changes in health behaviours score and in functional health, which is assumed to follow a Poisson distribution.

To test our hypotheses about the effect of person-centred and effective primary healthcare on changes in functional health, health behaviours and intention to change (objective 2), we will use Poisson or ordinal logistic regression. First, we will use separate regression models to estimate the effects of patient-centred primary healthcare at T1 on outcomes of functional health, chronic illness burden and health behaviours at T2 or T3. We will examine the effects of individual healthcare dimensions as well as global healthcare scores to better understand the relationships with outcomes. We will examine the presence of effect modification by multi-morbidity and of poverty by testing first-order interaction terms between healthcare and multi-morbidity/poverty in the regression model.

Second, we will use structural equation modelling and path analysis (LISREL) [82] to examine the relationships between the different dependent and independent variables (objective 3). For instance, we will test the paths by which chronic illness burden and health behaviours affect functional health, finding which variables mediate these relationships. We will look for the best explanatory model by comparing the Chi-square statistic of nested models as well as goodness of fit indices, such as the Comparative Fit Index (values of 0.90 indicate good fit) and the Root Mean Square Error of Approximation (RMSEA, values lower than 0.08 indicating acceptable fit) [83].

Sample size and statistical power

The sample size for this cohort is driven by the minimal size we need to detect a change in chronic illness burden and health behaviour change in 24 months. Estimates of incidence of chronic diseases vary by source, but in general we estimate that the annual incidence of having at least one of the physical or mental chronic illnesses of interest is approximately 100 per 1000. Assuming that incidence rates follow a Poisson distribution, a sample size of 2000 gives us 80% power to detect a rate difference of 18/1000 with α = 0.05 between any of our subgroups of interest. For path analysis, statistical power is a function of the number of variables in the model and the number of paths to be examined. Rule of thumb is that there should be 20 subjects per parameter. This sample size allows us to detect small size effects (β~0.15) in our paths of interest while controlling for individual variables.

Ethical considerations

Participation in the research has minimal risks. Major ethical concerns are ensuring confidentiality and maintaining participation throughout the study period. Nominal information will be stored separately from data, and only the project coordinator and principal investigator will have access to the link between nominal information and the unique study identification code.

The individual's right to withdraw partially or completely will be reiterated at each new data collection effort. The consent form, which explicitly states that the study is to be carried out over several years and consists of independent consents, was approved by the scientific and ethics committees of the Centre de santé et de services sociaux de Chicoutimi, as well as the Research Ethics Committee of Hôpital Charles Lemoyne.

Discussion

A study that follows the experience of a population sample over time will provide new and valuable information on the effectiveness of care in the population rather than in clients of selected care models. The study of how experience of primary healthcare evolves over time will be of specific value to decision-makers who implement system changes and will contribute to new knowledge in the area of measurement of healthcare experience. Focus on the patient's perspective is particularly relevant in an era of greater accountability to citizens, and reinforces the value base of primary care. Knowledge on the impact of introducing new models and on systemic effects of local configurations of healthcare and clinical governance in a population will shed new light on this issue. Repeated prospective measures provide richer information than a series of cross-sectional studies or retrospective designs. They will also generate new knowledge about the direction of relationships between care processes, patient evaluations, and individual characteristics, especially about how vulnerable persons navigate in the systems.

Strengths and limitations

A longitudinal cohort is vulnerable to selection bias through differential loss-to-follow-up. We will collect health and socio-demographic information at recruitment to assess the extent of differential loss-to-follow-up and will conduct sensitivity analysis to examine the impact of differential losses on inferences. Some volunteer bias is also likely to occur at recruitment, however, affecting population representativeness but not the validity of analytic inferences.

Response fatigue could lead to loss-to-follow-up and information error. However, response burden needs to be weighed against the strength of a cohort design that allows us to explore various outcomes over time gaining further specificity through repeated measures.

Overall, limitations and methodological challenges are far outweighed by the unique strengths of a longitudinal cohort. It is the only design that will provide the required information on the temporal direction of effects and explore a broad set of relationships. The focus on global illness burden and all types of first-contact access is not only highly relevant to primary healthcare practice and policy, but also allows us to detect important effects despite the modest cohort size.

Relevance and implications

To our knowledge this cohort is unique in Canada, and is also expected to yield results that are relevant internationally. Results will provide valuable information for primary healthcare clinicians on the course of chronic illness over time and the impact on health outcomes of accessible, patient-centred and effective care. A demonstration of impact will contribute to the promotion of continuous quality improvement activities at a clinical level. Finally, while considerable advances have been made in the management of specific chronic illnesses, this will make a unique contribution to effective care for persons with multiple morbidities.