Significant regional variation in prescribing quality was observed among older adults receiving primary care in VA. The frequency of inappropriate prescribing was highest in the South and lowest in the Northeast. This pattern mirrors a recent study among Medicare beneficiaries, using data from the same time period.22
Furthermore, this pattern has remained consistent over time. A study using 1999 data for community dwelling elders found older adults in the New England and Middle Atlantic census regions had significantly lower odds of inappropriate prescribing compared to the East South Central region.23
Similar regional patterns have been observed in the inpatient setting.24
The consistency of regional variation over time, across setting, between VA and non-VA systems, and among different quality measures, likely reflects differences in the general climate of clinical practice that extend beyond specific institutions or health care systems.
In addition to geographic region, prescribing quality varied significantly by rural residence. Potentially inappropriate prescribing was significantly more common among rural veterans across 4 independent quality indicators. However, the magnitude of the differences was likely to be of modest clinical significance. The largest difference was seen with Zhan criteria, where 18.7% of rural veterans received a potentially inappropriate medication, compared to 17.2% of urban veterans. Prior studies of rural-urban differences in prescribing quality have had mixed results, which could be explained by an apparently small effect size. 23,32,36,42,43
It is unlikely that rural veterans experience significant barriers to obtaining medication refills in VA due to the mail delivery system where refills can be requested in-person, by phone, or online. However, disparities in medication access could still arise from barriers to accessing the physicians and other providers who write prescriptions. This concern is the basis for most studies of rural-urban differences, where quality measures typically reflect the absence of some health care service. In contrast, all 4 measures of prescribing quality examined in this study reflect access to treatment that should not
have been received. Therefore, our findings cannot be attributed to impaired health care access, but perhaps they could be described as diminished access to high-quality care. For example, our findings could be explained by restricted access to specialty geriatrics care, given that 90% of geriatricians are located in urban areas.44
Looking beyond the national average, we further examined the impact of rural residence within specific geographic regions and documented a clinically significant interaction. Consistent with the national average, rurality was associated with greater rates of potentially inappropriate prescribing in the Northeast and the South. However, rural veterans were at significantly lower risk for inappropriate prescribing in the West, and they faced equivalent risk in the Midwest. This observation has important implications for future research. Most studies examine rurality as a single variable and assume a constant effect across geographic regions. Based on this assumption we would have concluded that rural veterans were at an increased risk for inappropriate prescribing. However, we discovered that rural residence could be associated with increased, decreased or equivalent risk, depending on the region. Therefore, national implementation of policy or interventions could be misguided and inefficient, when potential rural disparities are limited to specific geographic regions. Only one prior study has similarly examined regional differences in the impact of rural residence on health care quality.45
In parallel to our findings, this study found that rural individuals received lower quality diabetes care in the South but equivalent or higher quality care in other geographic regions. The authors attributed this finding to the higher proportion of non-whites in the South, but they could not provide direct analytic support for this hypothesis. The extent to which racial disparities could explain our findings is unknown, and it is an important avenue for future research.
There are several important limitations to consider. First, our study was limited to older adults, and variation patterns may be different for younger populations. While further examination of younger patient groups is warranted, older adults represent a key population of interest for prescribing quality. Second, our study was limited to veterans, and thus predominantly comprised men. However, a prior study of Medicare beneficiaries found similar regional patterns in prescribing quality during the same time period.22
Therefore, our findings do not seem to apply only to men, or to individuals receiving care in VA. A third limitation is the use of prescribing quality indicators, derived from electronic administrative data, as a measure of inappropriate prescribing. These indicators do not account for individual differences and thus reflect prescribing practices that can only be considered to be potentially inappropriate. While it has been argued that these indicators may have insufficient specificity for patient-level analyses,46
they can still serve as valid proxy measures of inappropriate prescribing at larger aggregate levels.47
As long as the positive predictive value does not vary systematically by geographic region, then variations in potentially
inappropriate prescribing will mirror underlying variations in actual
inappropriate prescribing. Furthermore, the observation of similar geographic variation patterns across 3 independent indicators supports the validity of our findings. The one indicator that did not follow the same pattern of regional variation was drug-drug interactions. It is noteworthy that VA employs drug-drug interaction screening software that is common across all VA medical centers, which could explain why less regional variation was observed with this indicator. However, there is still the opportunity for variation arising from differences in how clinicians respond to the warnings presented by the software. Finally, we were limited to prescription medications supplied through the VA and thus did not have access to information concerning medications acquired through other sources such as Medicare part D, private insurance, or purchased out-of-pocket.
In summary, we observed clinically meaningful regional variation in prescribing quality across VA, consistent with a prior report involving Medicare beneficiaries.22
This suggests that regional variation in prescribing quality is not a function of specific system-level characteristics of VA, and interventions designed to reduce variation may struggle against endemic regional differences in prescribing practices. On the other hand, the closed nature of the VA system and the standardized electronic medical record may make interventions to improve prescribing quality more feasible than in non-VA systems. Our second key finding was that rural veterans were at increased risk for inappropriate prescribing overall, but that this effect varied by region. This finding suggests that the impact of rurality on health care access and quality is complex, and that being a rural veteran in the West is different from being a rural veteran in the South. Further, the one-size-fits-all approaches to ensuring access to high-quality health care for rural veterans may not be effective, and in many regions such approaches may be unwarranted. While this study documented important variations in prescribing quality, the underlying factors driving these trends remain unknown, and they are a vital area for future research affecting older adults in both VA and non-VA health systems.