A Percutaneous Coronary Intervention Lab in Every Hospital?
Background—In 2001, 1176 US hospitals were capable of performing primary percutaneous coronary intervention (PCI), and 79% of the population lived within 60-minute ground transport of these hospitals. We compared these estimates with data from 2006 to explore how hospital PCI capability and population access have changed over time.
Methods and Results—We estimated the proportion of the population 18 years of age or older, living in 2006 within a 60-minute drive of a PCI-capable hospital, and we compared our estimate with a previously published report on 2001 data. Over the 5-year period, the number of PCI-capable hospitals grew from 1176 to 1695 hospitals, a relative increase of 44%; access to the procedure grew from 79.0% to 79.9% of the population, a relative increase of 1%.
Conclusions—Our data indicate a large increase in the number of hospitals capable of performing PCI from 2001 to 2006, but this increase was not associated with an appreciable change in the proportion of the population with access to the procedure. In the future, more attention is needed on changes in PCI capacity over time and on the effects of these changes on outcomes of interest such as service utilization, expenditures, patient outcomes, and population health.
For patients with ST-segment elevation–myocardial infarction (STEMI), percutaneous coronary intervention (PCI) is better than fibrinolytic therapy at reducing mortality1 if administered in a timely manner.2,3 However, PCI is available only at hospitals with cardiac catheterization labs, and fibrinolytic therapy remains the standard of care in the majority of US hospitals. The optimal distribution of PCI-capable hospitals has long been debated, and currently there is considerable interest in increasing access to the procedure for patients with STEMI.4–8
Editorial see p 9
A previous study estimated that 79% of the US population in 2001 lived within a 60-minute drive of a PCI-capable hospital.9 To explore how PCI capacity and access to the procedure have changed over time, we set out to update estimates to 2006. We also set out to make recommendations for sustaining a reliable and continual update of hospital PCI capability in the future.
Following methods described in previous work on drive times to US hospitals,9–13 we conducted a cross-sectional study of US hospital PCI capability and the access of the US adult population with PCI in 2006. Our data sources included the 2006 American Hospital Association (AHA) Annual Survey Database of all 50 states and the District of Columbia, the 2006 Health Care Cost and Utilization Project's (HCUP) State Inpatient Databases from 21 states, and the 2000 US Census. To meet inclusion criteria, a hospital had to provide acute care to the US adult (18 years of age or older) population. Government facilities, hospital units within an institution, psychiatric and drug dependency hospitals, long-term care facilities, and children's hospitals were excluded from the analysis. All eligible hospitals were uniquely identified through their AHA identification number and geo-coded within a geographic information system, using latitude and longitude coordinates. To assess the potential of PCI-capable hospitals to deliver timely PCI to patients with STEMI, we used Census data to estimate the percentage of the US population living within a 60-minute drive of one of these hospitals. We compared results from our cross section with results obtained in a previous cross section of 2001 data, matching our data and methods of counting hospitals and estimating drive times as closely as possible to those in the previous report.9 Further information on how our data and methods compare with the previous report is contained in online-only Data Supplement Appendix A.
The first step in the main analysis involved identifying hospitals that were capable of performing PCI in 2006. To do this, we combined reports of PCI laboratory ownership in the AHA survey with reports of PCI utilization in HCUP administrative data. The AHA survey identifies hospitals with the capability to perform adult interventional cardiac catheterization in all 50 states and Washington, DC; we supplemented AHA data with administrative discharge data from hospitals in 21 states reporting to HCUP, covering 100% of hospital discharges in 2006 for 49.98% of the adult US population. We identified hospitals offering primary PCI from ICD-9 codes (00.66, 36.06, 36.07, and 36.09) collected in HCUP's State Inpatient Databases. To be counted as PCI-capable in our study, a hospital had to perform at least 4 procedures over the course of the year. We therefore estimated the number of US hospitals with a PCI program in 2006 by identifying any hospital that reported ownership of an adult interventional cardiac catheterization laboratory or reported performing a minimum of 4 PCI procedures.
In 2001, 1176 of 4609 US hospitals (25%) had a primary percutaneous coronary intervention (PCI) program.
Hospital PCI capability in 2001 was sufficient to provide timely primary PCI to 79% of the population.
In 2006, 1695 of 4673 US hospitals (36%) had a primary PCI program.
Hospital PCI capability in 2006 was sufficient to provide timely primary PCI to 80% of the population.
From 2001 to 2006, hospital capability to perform PCI grew by 44%, whereas timely access to the procedure grew by only 1%. Thus, the expansion in hospital capability did not appreciably increase access to care.
We further used HCUP to distinguish between hospitals capable of performing PCI on a full-time (24 hours per day/7 days per week) basis from those capable of performing the procedure on a part-time basis (Monday through Friday, 7 am to 5 pm) to see if accounting for hours of laboratory operation changed the estimates of access to PCI. This was only possible in 18 states, where information on the time and day of procedure was available in HCUP. Hospitals that provided 4 or more PCI procedures during weekdays or weeknights and weekends were considered full-time; hospitals that provided 4 or more procedures during weekday operating hours only (Monday through Friday, 7 am through 5 pm) were considered part-time. In the remaining states, no estimates were available on operating times.
Survey responses about ownership of an adult interventional catheterization laboratory in the AHA survey were missing in 790 (16.9%) of 4675 eligible hospitals in 2006. To reduce the number of missing observations, we imputed PCI capability in 2006 if a hospital reported PCI capability in any of the previous 5 years (observation carried forward if a hospital reported PCI capability during 2001–2005), and we imputed no PCI capability if a hospital reported no PCI capability in 2007 or 2008 (observation carried backward if a hospital reported not having PCI capability during 2007–2008). The imputation procedure therefore carried reports forward or backward on the assumption that a PCI program would be continued once it was in place, given the large fixed costs involved in opening it. After imputation, 463 (9.9%) hospitals still had missing data. Missing observations were more likely to be found in the Northeast (12.4%) and West (12.2%) Census regions and least likely in the Midwest (6.3%).
The second step in our main analysis was to estimate the percentage of the adult population living within a 60-minute drive time to a PCI-capable hospital. We conducted a drive times analysis from populations (within Census tracts) to treatment access points (within hospitals), using network and speed limit data from the Environmental Systems Research Institute StreetMap USA (ArcMap version 9.2, Redlands, CA). Census tracts were divided into 3 groups: urban, suburban and rural, corresponding to tertiles of population density measured in people per square mile. We used these groups to adjust prehospital drive times by urban density. We started drive time estimates from Census tract centroids—the mean center of a tract—and followed the shortest time-distance to neighboring hospitals, recording all trips that could be made in less than 60 minutes. To estimate total elapsed time from 911 call to hospital arrival, we added extra time to account for dispatch of the emergency medical services (EMS) vehicle (1.4 minutes was added for urban and suburban tracts and 2.9 minutes for rural tracts), time from EMS depot to scene (total time was multiplied by a constant of 1.6, 1.5, or 1.4 for urban, suburban, or rural tracts, respectively), and time spent on scene (13.5 minutes was added for urban and suburban tracts and 15.1 minutes for rural tracts). These constants were derived in a meta-analysis of empirically determined prehospital care times for trauma.13 The sum of adjusted drive times, and these constants allowed us to identify all Census tracts and their populations with 60-minute, timely access to PCI.
The adult population of each Census tract whose centroid lay within 60 minutes of a PCI-capable hospital was considered to have timely access to that hospital. Populations in Census tracts covered by multiple hospitals were counted once to avoid duplication. In the 18 states that reported time and day of procedure, population counts surrounding hospitals with part-time PCI capability were weighted 0.39, representing the proportion of all heart attacks in the United States that begin during weekday operating hours.14,15 Tract-level data were aggregated at the state, regional (Midwest, Northeast, South, and West), and national levels in separate analyses.
The third major step in our project was to assess the validity of AHA survey data as the potential sole source of information on PCI capability and access to PCI in the United States. The AHA data are considerably less expensive and easier to obtain than Medicare claims and HCUP discharge data. To understand whether future drive times analyses might reliably depend on use of AHA data alone, we compared AHA reports with HCUP discharges in the 21 states reporting to HCUP in 2006. In this analysis, we used the 2 data sets independently to identify hospitals with PCI capability and to estimate timely access to PCI. We estimated the direction and extent of bias in AHA data, using HCUP administrative data as a benchmark.
All statistical analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC).
Hospitals invested heavily in new PCI capability during the 5 years from 2001 to 2006, making it newly available in 519 hospitals, for a total of 1695 (36.3% of all eligible hospitals) acute care hospitals in the United States This heavy investment in new PCI capability represented a relative increase of 44% over 2001 but did not result in an appreciable change in timely access to the procedure. As in 2001, approximately 80% of the population in 2006 lived within 60 minutes of a PCI-capable hospital. In 2006, 79.9% of the adult population in the 50 states and the District of Columbia (n=209 128 092) lived within 60 minutes of a PCI-capable hospital (Figure). In the 21 states for which we had information on the time and day of procedure, 84% of all hospitals were capable of providing the service on a full-time basis (24 hours per day, 7 days per week), and the remainder were capable of providing the service on a part-time basis only (Monday through Friday, 7 am to 5 pm). Accounting for the part-time operation of PCI labs reduced the level of access by 0.9 percentage points.
There was considerable regional variation in access to PCI in 2001, and variation persisted in 2006 (Table 1). Access to the procedure was highest in the Northeast (87.8%) and lowest in the South (75.7%). More than 90% of the population in 7 states and the District of Columbia had 60-minute access to primary PCI, including California (90.9%), Connecticut (93.6%), Delaware (91.7%), Florida (91.2%), Massachusetts (94.6%), New Jersey (96.5%), Rhode Island (96.1%), and Washington, DC (100.0%). Less than 50% of the population in 7 states had 60-minute access to the procedure, including North Dakota (48.9%), South Dakota (40.3%), Vermont (38.3%), West Virginia (45.6%), Alaska (40.0%), Montana (45.3%), and Wyoming (30.5%).
For a subset of the population, we saw mixed evidence of improvement in potential times to hospital arrival. The percentage of the population living closest to a PCI-capable hospital improved from 42% to more than 51% in 5 years' time, potentially reducing drive times to primary PCI for at least 9% of the population. For the nation as a whole, however, this benefit did not appear to make much difference for drive times. In 2001, projected median drive time nationally was 11.3 minutes, and this figure dropped 10.5 minutes (interquartile range, 6.2–18.0) 5 years later, an improvement of only 48 seconds for the typical patient (Table 1). Projected median elapsed time from 911 call to arrival at the closest PCI hospital in 2006 was 25.6 minutes (interquartile range, 21.2–33.8) (Table 1), a drop of only 30 seconds compared with 2001.
Our primary purpose in this study was to estimate access to PCI labs in 2006 and to compare this with a previously reported estimate from 2001. Our secondary purpose was to assess whether inexpensive, readily available AHA survey data could be validated as the sole source of information on PCI capability in future research. To accomplish this, we conducted an analysis comparing AHA survey data with HCUP administrative discharge data. This subanalysis showed little meaningful difference between self-reports and empirical reports of PCI capability (Table 2). In self-reported survey data, we found a small negative bias in the count of hospitals with PCI capability and small positive bias in timely access to the procedure. Using HCUP data alone in 21 states, we identified 710 hospitals that were capable of performing PCI and estimated that 81.7% of the population had timely access. Using AHA data alone in the same 21 states, we identified 696 PCI-capable hospitals and estimated that 83.2% of the population had timely access to PCI, an absolute difference of only 1.5% in access to the procedure. The gap between self-reported and empirical data on access to PCI was largest in North Carolina (approximately 10 percentage points; a relative difference of 17%), but in most places it was very small.
We used a κ statistic to measure the extent of agreement between the 2 data sources, where perfect agreement is indicated by κ=1 and perfect independence is indicated by κ=0. We used it to test a null hypothesis of independence in the classification of PCI capability, after using the 2 data sets as alternative sources of information. For the 21 states in which we had both data sources, the summary κ statistic was 0.72, representing strong agreement between the 2 data sets. In 7 states, agreement was excellent, above 0.80, and in 3 states agreement was moderate, at 0.47–0.60. There were 137 (6.5%) disagreements; in 82 cases AHA data reported ownership where our HCUP criteria did not indicate ownership, and 55 hospitals reported not owning PCI, whereas HCUP data recorded at least 4 procedures during the year. In every state but Vermont, where only 2 PCI labs were found (P<0.011), confidence to reject independence was well above 99%.
This study established several important new findings about hospital PCI capacity in the United States. Our data show that hospitals have invested heavily in new PCI programs over a 5-year period, without producing an appreciable change in access to the procedure. We showed that hospital capability to perform PCI grew by 44% from 2001 to 2006, whereas access to the procedure held steady at approximately 80% of the population. We showed continued variation in access to these centers from region to region. We showed that increased PCI capacity is associated with only modest reductions in projected time to care. Finally, we found that bias in AHA surveys is small, suggesting that these data may be suitable as the sole source of information on hospital PCI capability in future research on hospital markets, patient outcomes, treatment costs, regional planning, and regulatory policy.
Our findings suggest that the capacity of US hospitals to perform PCI may have been understudied since the last published report appeared on 2001 data. There are several urgent research questions that arise from our results. First, more information is needed about the hospitals and markets in which new PCI capability has been introduced in recent years. Are new programs opened in for profit or nonprofit hospitals, in larger or smaller ones? Are new PCI programs more or less likely in neighborhoods where a PCI program already exists? Recent evidence in Michigan showed that new PCI programs were more likely to open in hospitals without coronary artery bypass graft backup.16 Other research has shown that new coronary artery bypass graft programs are more likely to be duplicative of older programs.17 Research is urgently needed to learn more about the hospitals where PCI programs are being opened.
More information is also needed on the relationship between changes to PCI capability over time and outcomes of interest such as service utilization, expenditures, patient outcomes, and population health. We currently know little about how recent changes to PCI capacity are related to changes in care and population health. Recent research shows that PCI utilization rates have remained flat since 2001.18 How are utilization rates changing for the subset of patients with STEMI over this period? Has the proliferation of PCI labs improved outcomes through the reduction of time to treatment, or has it worsened them through reductions in procedure volumes? Likewise, we know little about the effects of PCI diffusion on the costs of care. Although it is widely agreed that technology change accounts for a large share of growth in health spending,19 very little is known about how recent expansions of coronary revascularization programs affect health expenditures overall.
Finally, better information is needed on how PCI capability has changed in small regions, to shed light on the optimal strategies for both STEMI regionalization efforts and regulatory policy on PCI. To address the need for better access to PCI, many have advocated for a system of voluntary regionalization. The American Heart Association's Mission: Lifeline program,20 for instance, identifies 410 communities that have implemented some form of regionalized PCI strategy. Strategies to achieve regionalized care include community-wide planning for distribution of PCI capability, 911 programs for detecting and diverting patients directly to PCI hospitals, and regional agreements between hospitals to select high-benefit patients for interhospital transfer.6,21 Previous studies comparing local strategies for STEMI regionalization have suggested that the construction and staffing of new PCI labs is more costly and less effective than a strategy of diverting patients to already-existing PCI hospitals.22,23 Could expansion of new PCI programs be optimized to increase access to the procedure? Similarly, better information is needed to evaluate existing state health planning activities. American Health Planning Association annual reports show that roughly half of states continue to support certificate of need programs with specific provisions for cardiac catheterization services review,24–26 suggesting that public policy remains unsettled on whether regulation should be used to restrain the diffusion of PCI. Should regional planners emphasize new PCI program restrictions, investments in better 911 and EMS infrastructure, or a combination of strategies?
This study faced some limitations. First, hospital discharge files from HCUP were available for only 21 states and we were only able to adjust population estimates for part-time PCI capability in 18 states that collect data for day and time of procedures. To address these constraints, we compared the 2 data sources in states covering nearly 50% of the US population 18 years of age and older. This comparison validated AHA data as the sole source of information on hospital PCI capability, and we are confident that the addition of a nationwide administrative data set would not add much value to the analysis. Second, our analysis focused on elapsed time between 911 call and hospital arrival, and this time segment is just 1 factor affecting timely treatment for patients with STEMI. Patient awareness of symptoms, mode of transport, 911 system capacities and response times, hospital door-to-balloon times, and other factors play a role. Third, we lacked patient level outcomes data and therefore could not assess a range of questions that arise immediately after absorbing the findings of our study. Our findings underscore the importance of further research on the hospitals in which new PCI programs are introduced, the clinical outcomes of patients surrounding these programs, the impact of new programs on costs, and the role of voluntary regionalization and regulation strategies on these programs.
The last published report of hospital PCI capability was based on 2001 data. Our findings suggest that more frequent assessments of hospital PCI capacity are urgently needed. We laid the groundwork for conducting such assessments, using low-cost, readily available AHA surveys as the sole source of information on hospital PCI capability. We identified a broad range of research questions that could be answered with longitudinal updates, including evaluations of hospital-level and other factors that influence the adoption of PCI, assessments of the impact of new programs on patient outcomes and health care expenditures, and evaluations of STEMI regionalization programs and regulatory policy.
Sources of Funding
Dr Concannon was supported by the Agency for Healthcare Research and Quality (K01 HS017726) and by the Tufts Medical Center Research Fund. Jason Nelson was supported by the Tufts Medical Center Research Fund and the Tufts Clinical and Translational Science Institute, Award No. UL1 RR025752 from the National Center for Research Resources, National Institutes of Health. Jessica Goetz was supported by the Agency for Healthcare Research and Quality (K01 HS017726). Dr Griffith was supported by the Tufts Clinical and Translational Science Institute, Award No. UL1 RR025752 from the National Center for Research Resources, National Institutes of Health.
The online-only Data Supplement is available with this article at http://circoutcomes.ahajournals.org/lookup/suppl/doi:10.1161/CIRCOUTCOMES.111.963868/-/DC1.
- Received March 8, 2011.
- Accepted October 24, 2011.
- © 2012 American Heart Association, Inc.
- Jacobs AK
- Jacobs AK,
- Antman EM,
- Faxon DP,
- Gregory T,
- Solis P
- Jacobs AK,
- Antman EM,
- Ellrodt G,
- Faxon DP,
- Gregory T,
- Mensah GA,
- Moyer P,
- Ornato J,
- Peterson ED,
- Sadwin L,
- Smith SC
- Topol EJ,
- Kereiakes DJ
- Nallamothu BK,
- Bates ER,
- Wang Y,
- Bradley EH,
- Krumholz HM
- Willich SN,
- Lowel H,
- Lewis M,
- Hormann A,
- Arntz HR,
- Keil U
- Lucas FL,
- Siewers A,
- Goodman DC,
- Wang D,
- Wennberg DE
- Smith S,
- Newhouse JP,
- Freeland MS
American Heart Association. Mission: Lifeline. http://missionlifelinecommunity.americanheart.org/home. Accessed July 28, 2010.
- Concannon TW,
- Kent DM,
- Normand SL,
- Newhouse JP,
- Griffith JL,
- Cohen JT Beshansky JR,
- Wong JB,
- Aversano T,
- Selker HP
National Directory of Health Planning, Policy and Regulatory Agencies. 5th edition. Falls Church, VA: American Health Planning Association; 1993.
National Directory of Health Planning, Policy and Regulatory Agencies. 12th edition. Falls Church, VA: American Health Planning Association; 2001.
National Directory of State Certificate of Need Programs and Health Planning Agencies. 19th edition. Falls Church, VA: American Health Planning Association; 2008.