Usage of a health information exchange system at the point of care reduced the occurrence of repeat imaging procedures in a multi-payer community sample.
Published Online: January 14, 2015
Joshua R. Vest, PhD, MPH; Rainu Kaushal, MD, MPH; Michael D. Silver, MS; Keith Hentel, MD, MS; and Lisa M. Kern, MD
We modeled the binary outcome of repeat imaging using a binary logit model with generalized estimating equations. We chose this method because it accounts for the clustering that occurs with repeated measures. With the exception of the patient, we treated all other measures as fixed effects. We exponentiated parameter coefficients to express odds ratios (ORs). We adjusted for the following clinically relevant patient-level variables: payer type, age, gender, number of primary care visits, number of specialty care visits, number of ED/urgent care visits, number of admissions, and the count of major ADGs. We conducted stratified analyses for the 3 modalities with sufficient sample sizes: CT, ultrasound, and radiographs.
Frequency of Medical Imaging and Repeat Imaging
The entire cohort consisted of 196,314 patients. The first 3 months of claims for the patient cohort included 68,296 claims for imaging procedures. After de-duplication (removing multiple claims associated with the same imaging procedure and those that had no body region identified), we were left with 56,306 imaging procedures. Overall, 17.6% (n = 34,604) of patients had at least 1 imaging procedure, equivalent to a rate of 28.7 imaging procedures per 100 patients. As displayed in Table 1, the most common imaging modalities were radiographs (43.7% of all imaging procedures), CT (16.4%), ultrasound (16.4%), mammography (10.0%), and magnetic resonance imaging (MRI) (6.8%). Although we considered 23 different modalities, these 5 accounted for more than 90% of all imaging procedures.
Overall, 7.7% of medical imaging procedures (n = 4316) were repeated within 90 days. As indicated in Table 1, that percentage varied widely by modality. The rate of repeat imaging was highest for ultrasounds: 15.5%. Similarly, 8.6% of radiographs, 4.7% of mammograms, 3.8% of CTs, and 2.5% of MRIs were repeated. Other procedures with high rates of repeats included echocardiography (12.5%) and urography (7.0%), although the absolute numbers of these tests were lower. For 2 other modalities, less than 1% each were repeated; another 8 modalities had no repeat imaging. The timing of the repeat imaging occurred at various points over the course of the 90-day follow-up period (Figure). Overall, half of the repeated procedures occurred within the first 30 days after the initial medical image. By 60 days, a total of 80% of the repeated procedures had occurred.
The majority of the imaging procedures (73%) were among women and among the privately insured (59%, Table 2). The average patient age was 57.2 years, and the average number of total healthcare encounters in the 90 days after the initial imaging procedure was 2.7.
Use of the HIE System
Overall, providers accessed the HIE system within 90 days after 11.8% of imaging procedures. As displayed in Table 2, when providers accessed the HIE system, those procedures were more likely to be for patients in Medicare managed care, for older patients, and for sicker patients, with higher counts of major ADGs. In addition, accessing the HIE system was more likely to occur with higher numbers of healthcare encounters.
Association Between HIE Usage and Repeat Medical Imaging
We found that if the HIE system was accessed within the 90 days following an initial imaging procedure, the imaging was less likely to be repeated (5.2% of imaging procedures were repeated when the HIE system was accessed versus 8.0% repeated when the HIE system was not accessed). The unadjusted odds of repeat imaging were 44% lower if the HIE system was accessed after the initial procedure (OR = 0.56; 95% CI, 0.49-0.65; Table 3). After controlling for patient characteristics and utilization, provider access of the HIE system after the initial imaging was independently associated with 25% lower odds of repeat imaging (OR = 0.75; 95% CI, 0.65-0.87). Given the rate of repeated imaging observed in this population (7.7%), out of every 36 images, HIE access would prevent 1 repeated image that would have occurred otherwise.
These results persisted when we considered ultrasounds alone and radiographs alone (Table 4). Provider access of the HIE system reduced the adjusted odds of both a repeat ultrasound by 44% and a repeat radiograph by 21%. HIE usage was not associated with the odds of a repeat CT, although the sample size of repeat CTs was much smaller than that of the other 2 tests, limiting power for that comparison.
In our community-based study, we found a rate of 28.7 imaging procedures completed for every 100 patients over a 6-month period. We also found that 7.7% of all imaging procedures were repeated within 90 days. When repeat imaging was done, it tended to occur quickly, with nearly 50% of all repeated imaging completed within 30 days and 80% completed by 60 days.
We found that if the community-based HIE system was accessed by providers within the 90 days following an initial imaging procedure, the imaging was significantly less likely to be repeated, with 5.2% of imaging procedures repeated when the HIE system was accessed, compared with 8% repeated when the HIE system was not accessed. Adjusting for potential confounders, the odds that an imaging procedure was repeated decreased by 25% with HIE access.
This study provides one of the few estimates of the frequency of repeat imaging for multiple modalities in a multi-payer, multi-provider community. Our finding of a 7.7% rate of repeat imaging is slightly lower than rates found by other investigators, such as 9%, 13%, and 20%.8,9,29 Unlike previous studies, our study included multiple settings of care and a broad, community-based patient population. Previous work supports this distinction, as imaging is overall less frequent in the ambulatory setting than in the inpatient or ED settings.30-32 Additionally, several previous estimates were derived from patient populations in which multiple and frequent imaging is to be expected (eg, trauma or neurological). Our sample is likely more representative of the overall adult population. There are few interventions that have been tried for reducing the frequency of medical imaging. Payers have tried prior authorization for certain imaging procedures, but it is not always clear that the cost of the prior authorization program is outweighed by savings from fewer images.33 Other interventions that have been tried are conceptually close to information exchange, such as electronic decision support for ordering physicians (which often includes access to prior results) and picture archiving and communication systems (PACS) for electronic sharing of actual images.34,35 Electronic decision support for ordering imaging is still an emerging tool, and PACS systems have typically been installed within a single institution. Other approaches to image sharing, like digital media transfers, may also be effective, but those approaches can be cumbersome and tend not to include breadth of clinical data about the patient, as is found in the Rochester RHIO system.4,7,36-38
The existing literature on the effects of HIE on patient healthcare utilization, in general, is sparse,39 and the few studies that examine the effects on imaging do not present a consistent picture. A series of studies among ED patients at one RHIO reported similar reductions in imaging utilization for select patients, modalities, and locations when an HIE system was utilized at the point of care.14,16,17 Other reports of information-sharing technology also suggest that reductions in repeated and overall imaging usage are possible.40,41 Conversely, other examinations have found that adoption of exchange-capable health information technology is not associated with reductions in the rates of imaging ordering.18,19 However, those studies differ from this investigation, as they did not measure actual usage of the system.
Our study has several limitations. First, from our secondary sources, we could not determine the appropriateness of the imaging procedures. Our study measured all repeat imaging observed; we were not able to distinguish between procedures that were clinically appropriate and those that were potentially unnecessary. Some of the repeated procedures are clearly clinically appropriate and expected. For example, clinicians may need to determine changes in status or decide if new interventions are warranted. Further research could move toward separating the potentially unnecessary from the potentially appropriate imaging. Second, we were not able to adjust for all potential confounders at the provider level due to the fact that claims data do not consistently include the ordering provider. However, we tried to overcome this limitation through our procedure-level analysis, and it is likely that the same providers accessed the HIE for some of their patients and not for others; this would minimize the impact of any provider variables. Third, we know that providers accessed the HIE, but we cannot tell which particular data element may have affected their medical decision making. Understanding which pieces of information influenced changes to decision making would require alternative study designs.
PDF is available on the last page.