Analysis informing policy-making on the availability, accessibility, and utilization of primary medical care in rural settings requires data on consumers, providers, and the healthcare system summarized at the level of the physician service area (PSA). Through including/excluding residents and physicians, the specification of rural markets affects the quantification of variables estimating utilization and its determinants.
PSAs typically consist of one or more sites containing physicians (city, town and/or village) plus the surrounding hinterland (smaller villages and more sparsely settled areas) whose residents obtain medical care from providers. The approach developed here constructs PSAs (economic markets) by satisfying two rules: i) most residents obtain care from in-area physicians while ii) few visits are rendered to out-of-area patients (see Morrisey 1991). Because of their key role in the delivery of medical care to surrounding (and beyond) rural areas, influential towns, and the two cities were also identified as PSAs. The underlying philosophy is that out-of-area and in-area patterns of utilization are distinct processes that should be distinguished to understand rural issues better.
PSAs were constructed by testing alternative combinations of rural municipalities for the proportion of ambulatory visits provided to (by) area residents (physicians) by (to) in-area physicians (out-of-area residents). The final grouping was selected based on the best overall configuration defined as the minimum error from misclassifying patterns of utilization. From the population perspective, misclassification was defined as the percentage of all ambulatory visits made to out-of-area physicians while, from the physician-supply perspective, it was defined as the percentage of the workload rendered to out-of-area residents.
STEP 1: Count primary care ambulatory visits utilized in all municipalities. The measure of medical care utilization used to construct physician service areas is the ambulatory visit. This definition, which encompasses all visits made to physicians while the patient is not an inpatient (Tataryn et al.1995),includes patient--physician interactions occurring in the physician's office, the patient's home (including long-term care facilities), and hospital outpatient clinics/emergency room. Unless otherwise indicated, ambulatory visits include consultative care. Ambulatory care rendered as part of global tariffs (e.g., postoperative care, prenatal/postpartum care) is excluded. The ambulatory visit is the foundation for all subsequent medical care utilization -- both during contact (e.g., diagnostic and therapeutic services) and beyond (e.g., referral to a specialist, admission to hospital for medical or surgical care, follow-up ambulatory visits). Ambulatory visits made by rural residents during the fiscal years 1990-93 was used to identify the patient's place of residence and the general practitioner's location of practice (see Wall and Bogdanovic
Because of high physician-turnover in rural settings, particular attention was paid to tracking physician movement to assign visits to PSAs accurately. This three-year period yields stable estimates of patterns of utilization. Patterns estimated for shorter periods may be unduly influenced by gaps in practice-coverage during periods of physician-turnover while longer time-series increasingly incorporate conditions no longer relevant to the current (and future) situation. Because more than 90% of Manitobans contact a physician at least once within a two year period (Tataryn el aL 1995), these patterns are stable and, so, unlikely to change over the study.
STEP 2: Combine municipalities to construct the preliminary PSAs. The small areas defined by Zajac (1991) provide a starting point for grouping rural municipalities into physician service areas. Southern Manitoba (everything south of the 53rd parallel of latitude) was partitioned into rural, rural influential and urban areas (see Figure 3. 1). The two cities and certain rural towns (characterized by large concentrations of physicians and hospital beds providing more than 100,000 visits over the three years) were collectively identified as influential communities -- that is, communities in which patients from adjacent (and more distant) areas obtain medical care. The city of Winnipeg exhibits a province-wide influence on patterns of careseeking. Both the city of Brandon and the town of Portage La Prairie (CP) provides care within their respective regions. The towns of Morden-Winkler (CM), Steinbach (ES), Selkirk (IS), Dauphin (PD), Swan River (PS) and Neepawa (WN) exert important, but more localized, effects on patterns of care-seeking.
As the Zajac small areas were not adjusted for actual patterns of utilization, however, rural municipalities were regrouped to ensure that most visits occurred within these PSAs. From the population (physician-supply) perspective, goodness-of-fit was assessed as the proportion of ambulatory visits made to in-area physicians (proportion of ambulatory visits rendered to in-area patients). Although goodness-of-fit assessment found that these groupings satisfied the plurality rule, substantial misclassification (i.e., greater than 10%) occurred from both care-seeking and care-giving crossing boundaries between adjacent rural areas (see Wall and Bogdanovic 1995).
STEP 3: Combine/reassign municipalities to form larger PSAs. Building on Step 2, alternative combinations of rural municipalities were considered to represent patterns of both population care-seeking better and physician care-giving. First, three influential areas were reduced in size by reassigning municipalities with at least 10% out-of-area care-seeking to the appropriate rural PSAs. Next, alternative combinations of selected rural municipalities (high out-of-area care-seeking and/or care-giving) were tested. Overall, substantial improvement was afforded through these revised areas. See Figure 3.1 for the final spatial configuration of rural PSAs.
From both the care-seeking
and care-giving perspectives, misclassification error was reduced to less
than 10% (see Wall and Bogdanovic 1994), except 14 (11.04% for care-seeking);
however, merging 14 with 13 would create an excessively large PSA. From
the physician perspective, while some care-seeking misclassification remained,
most error was less than 10%. In drawing patients from more distant rural
settings, the influential PSAs were expected to exhibit high out-of-area
care-giving. Only Selkirk (IS) exhibited a low value (less than 30%) for
which little improvement would be derived by combining adjoining PSAs.
As expected, the combinations of municipalities forming these PSAs clearly
reflect the underlying influence of the rural highway system on patterns
of primary medical care utilization.