Impact of family practice continuity of care on unplanned hospital use for people with serious mental illness

Abstract Objective To investigate whether continuity of care in family practice reduces unplanned hospital use for people with serious mental illness (SMI). Data Sources Linked administrative data on family practice and hospital utilization by people with SMI in England, 2007‐2014. Study Design This observational cohort study used discrete‐time survival analysis to investigate the relationship between continuity of care in family practice and unplanned hospital use: emergency department (ED) presentations, and unplanned admissions for SMI and ambulatory care‐sensitive conditions (ACSC). The analysis distinguishes between relational continuity and management/ informational continuity (as captured by care plans) and accounts for unobserved confounding by examining deviation from long‐term averages. Data Collection/Extraction Methods Individual‐level family practice administrative data linked to hospital administrative data. Principal Findings Higher relational continuity was associated with 8‐11 percent lower risk of ED presentation and 23‐27 percent lower risk of ACSC admissions. Care plans were associated with 29 percent lower risk of ED presentation, 39 percent lower risk of SMI admissions, and 32 percent lower risk of ACSC admissions. Conclusions Family practice continuity of care can reduce unplanned hospital use for physical and mental health of people with SMI.


| INTRODUC TI ON
Serious mental illness (SMI) includes schizophrenia, schizoaffective disorder, bipolar disorder, and other psychoses. People with SMI have high rates of comorbidity, 1 reduced quality of life, 2 shortened life expectancy, 3,4 and high rates of emergency department (ED) presentations and unplanned hospital admissions. [5][6][7] Finding ways to improve health care and outcomes for this group is therefore a high priority. 8 Continuity of care is widely held to be beneficial for people with long-term conditions, including SMI. It is valued by patients 9,10 and providers 11 and considered good practice in mental health and family medicine, [12][13][14] reducing fragmentation of care and facilitating better provider-patient relationships. 15 Relational continuity-the longitudinal relationship between a patient and a health care practitioner (or group of practitioners) 16 -is often the focus of efforts to improve continuity. To date, evidence has been mixed on whether relational continuity improves outcomes for people with SMI. Some studies have found that higher continuity is associated with lower mortality, 17 reduced hospital admissions, 18 and improved recovery from episodes of SMI, 19 while others have found no association or even the reverse. [20][21][22] Studies that have examined the relationship of continuity to costs have mostly found that higher continuity was associated with lower health care costs, although one showed an association with higher costs of community care 18,23,24 It is important to clarify whether relational continuity is beneficial, since achieving higher continuity may increase costs and require trade-offs with other elements of good care, such as flexibility to meet urgent care needs. 25 Studies of relational continuity for people with SMI have most often considered visits within specialist mental health services, or across multiple types of service 21 (which we term "across-practice continuity"). However, in the UK family physicians provide much of the physical and mental health care for people with SMI and around a third of people with SMI are treated solely by their family physician. 26 Policies such as named accountable practitioners have emphasized the importance of maintaining continuity with an individual family physician, not just a practice. 27 The UK's National Health Service (NHS) provides publicly funded health care which requires patients to register with a specific family practice, so that patients face barriers to changing practices or attending different practices concurrently. In other health care systems, the role of family physicians in the care of people with SMI may be less prominent, 28 and patients may be more likely to see physicians at different family practices, but initiatives such as the patient-centered medical home in the United States have a similar focus on relational continuity with family physicians. 29 Evidence is therefore needed on the impact of within-practice family physician continuity on the physical and mental health of people with SMI, in addition to the existing literature on across-practice continuity focused on specialist mental health care.
Continuity of care has other aspects beyond relational continuity, including informational and management continuity. 16 In the United Kingdom, people usually register with a family practice and within that practice have a nominated physician who acts as a gatekeeper to and liaison with other health care services, including specialist mental health services. However, individuals can see any physician in that practice, especially for urgent appointments. Care plans for people with SMI document the patient's care needs, patterns of relapse, preferences for treatment, and social context 30 and are stored with patient records and accessible by different practitioners seeing the patient. Care plans therefore promote informational continuity across family physicians in the same practice and may also promote management continuity, if the management approach is agreed and can be followed by all practitioners. A previous study showed that care plans for people with SMI were associated with a lower risk of unplanned hospital use, but that study did not account for relational continuity. 31 Relational continuity is known to vary with observed individual characteristics such as age and sex, [32][33][34] but continuity may also be influenced by factors that are usually unobserved, such as helpseeking attitudes, disease severity, personality, or social context.
If these unobserved factors also influence outcomes, the observed association between continuity and outcomes may be biased. For example, people who are more proactive in seeking care may receive higher continuity, but they may also have better outcomes because they seek care early or engage in preventive management.
Conversely, family physicians may prioritize continuity for people with more severe illness, who nonetheless may have a higher risk of deterioration than those with less severe illness. To our knowledge, only one study has attempted to address unobserved confounding when examining the relationship between continuity of care and outcomes. It looked at the effect of relational continuity on emergency department attendance for people with diabetes and hypertension in Taiwan and measured continuity in 1 year and outcomes in the next. 35 It employed an instrumental variable approach to account for confounding, with the relational continuity of family members of the patient as instrument. The results showed a stronger negative association between continuity and ED presentations with the instrumental variable approach than the standard approach.
We examined whether family physician relational continuity for people with SMI is associated with better outcomes, using the novel application of methods to account for time-invariant unobserved confounding. The study objective was to investigate the hypothesis that continuity of care in family practice reduces unplanned hospital utilization.

| Study design
This observational cohort study used individual-level family practice administrative data linked to hospital administrative data to investigate the relationship between family practice continuity of care for people with SMI and time to unplanned hospital use.

| Sample
We used data from the Clinical Practice Research Datalink (CPRD), a database of anonymized patient records derived from over 600 RIDE Et al. family practices in England and broadly representative of the national population with respect to age and gender. 36 The records were linked to Hospital Episode Statistics (HES), which capture all hospital admissions (for both physical and mental health) and ED presentations funded by the NHS. This covers the majority of these types of health care in England, since the NHS funds 88 percent of all health care expenditure 37 and 92 percent of hospital care, 38 and there are no privately funded emergency departments. The sample was all people with a diagnosis of SMI documented in primary care on or before March 31, 2014 (the end of the study period), whose records met CPRD quality standards, and who were registered during this period at a participating practice that met CPRD standards. 36 Diagnoses of SMI were based on clinical information in routine practice data recorded in Read codes, an hierarchical coding system for clinical data that classifies diseases, patient characteristics, tests, and procedures 39,40 (see Table S1 for a list of the Read codes used in this study).    Table S2.) All ED presentations were included. For each type of outcome, we considered only the first observed instance (presentation or admission), since this could have influenced subsequent continuity.

| Outcome measures
The occurrence of the outcome is measured in the 3-month period t and continuity is measured over a lookback period of the prior 12 months (4 × 3 month periods t−4 to t−1). That is, there is no overlap between the 12-month period in which continuity is measured and the subsequent 3-month period in which outcomes are observed.
The outcome variable is a binary variable for each 3-month period indicating whether or not the event occurred in that period. For any individual who did not experience the outcome of interest (eg someone who did not present to ED during the period of observation), this variable is equal to zero for all periods. As we only analyzed time to first event, for any individual who did experience the outcome, the variable is equal to zero for all periods except the final period and equal to one for the final period, with all periods after the first event excluded from the analysis for that outcome.  of visits which are to the same family physician out of the total number of consecutive pairs of visits at the practice. Each index ranges from zero (lowest continuity) to 1 (perfect continuity). Tables S1-S6, and illustrative examples are shown in Table 1.

Additional detail on each index is in the
We measured continuity over 12 months (4 × 3 month periods), considering only face-to-face visits with family physicians. There is no standard level for "high" and "low" continuity, so we applied one recognized method that classified relational continuity as "high" if the level of continuity was above the median for the index, and "low" if at or below the median level. 43,44 A minimum of two visits is required to calculate COC and SECON, but to improve index stability we set the minimum to three visits.
Periods with fewer than two visits in the prior 12-month lookback period were included in the analysis with continuity categorized as "undefined." We constructed a set of categorical variables based on visit frequency and whether continuity was low or high. This allowed for different effects of continuity for frequent and less frequent users of family practice, as suggested by previous research. 45 Visit frequency was classified into low, moderate, and high: low (0-2 visits), moderate (3-5 visits), and high (6 or  Periods were then classified into five categories according to continuity level and visit frequency in the prior 12 months: low visit frequency (with continuity undefined-the base category), moderate visit frequency with low continuity, moderate visit frequency with high continuity, high visit frequency with low continuity, and high visit frequency with high continuity.

| Statistical analysis
The To allow comparison of our results to previous studies examining the effect of continuity of care, we also estimated models that did not specify individual heterogeneity as a function of the means of the time-varying variables, the random-effects model. These models allow for normally distributed individual heterogeneity but it is assumed to be uncorrelated with the explanatory variables contained in the model.
All models included observed individual characteristics as explanatory variables and adjusted standard errors for clustering at the practice level. We estimated separate models for each of the three continuity indices because of multicollinearity of the indices.
All analyses were conducted using Stata v14.

| Robustness checks
1. We tested the sensitivity of the results to the level of visit frequency at which continuity was classified as "undefined".
The minimum level for measuring continuity (and corresponding categories for low vs moderate visit frequency) was set to two or four visits rather than three visits as in our main analysis.

| Sample
The sample consisted of 19 324 individuals attending 215 practices, observed for 15.8 3-month periods on average (range 1-28 periods).  The standard approach (random effects) to modeling continuity, which does not account for unobserved confounding, produced different results, especially regarding care plans, as seen in the final column of Table 3. This approach found that higher relational continuity was associated with lower risk of ED presentation and lower risk of ACSC admission, at both moderate and high visit frequency, and that care plans were associated with higher rather than lower risk of SMI admission.

| D ISCUSS I ON
We found that within-practice family physician relational continuity for people with SMI was associated with a lower risk of ED presentations and ACSC admissions, and all-cause unplanned admissions.
These effects were present after accounting for time-invariant confounding, and across three dimensions of relational continuity as captured by three different continuity indices. We did not find significant association between relational continuity and risk of SMI admission.
Consistent with a previous study of care plans in family practice for people with SMI, 31 we found that care plans, which may represent informational/ management continuity, were associated with lower risk of ED presentations, but unlike that study (which did not account for time-invariant confounding), we found that care plans were also associated with lower risk of SMI admissions. We also found care plans were associated with lower risk of ACSC admissions.
Our results suggest that seeing the same family physician over time can improve the physical health of people with SMI and thereby reduce their need for and use of unplanned hospital care. These findings are consistent with previous studies that found relational continuity to be associated with reduced risk of ACSC admission in a range of different patient groups. 45,52 Higher continuity of family physician care may reduce the need for hospital care through improved management of physical health, by facilitating familiarity, communication, trust, and quality of relationship between doctor and patient. 15 The results also suggest that the documentation and sharing of information and management plans across physicians within a family practice can have important benefits for both the physical and mental health of people with SMI. Documentation of care plans was associated with reduced risk of all types of unplanned hospital care.
Our results also highlight the importance of accounting for the individual's propensity to receive continuity of care when studying the impact on outcomes. We used a modeling technique, the correlated

| Limitations
The clinical outcomes we have examined are important as they rep-

RIDE Et al.
time-invariant unobserved individual characteristics, we cannot rule out time-varying confounding that may contribute to our findings. For instance, during periods of deterioration leading to admission, family physicians may have less opportunity to spend time on preventive measures such as care plans. We were unable to differentiate the nature of ED presentations into physical and mental health as done for admissions because this level of detail was not sufficiently recorded in the original data. Care in specialist mental health services might be expected to confound the relationship between continuity in family practice and hospital use. However, we found that although specialist care was strongly associated with higher risk of each outcome, there was no change in the associations between the continuity and each outcome. While this was tested on a smaller sample with a shorter observation period due to data constraints, it provides reassurance that our main results are not biased by the absence of specialist mental health care in the model.

| CON CLUS IONS
Our results suggest that continuity of care in family practice, in