The VHA has been highly successful at improving blood pressure (BP) control, exceeding performance on existing BP measures by 79% versus only 62% in Medicare. Older individuals stand to benefit from good BP control, especially in stroke and myocardial infarction reduction. However, older individuals are at risk of falls due to multiple medical conditions and polypharmacy. One concern is that our existing dichotomous BP targets result in inadvertently-low BPs, especially as aging Veterans develop geriatric conditions such as fall risk. We have previously found that nearly one-third of older Veterans with diabetes are potentially-over-treated. Whether or not VHA providers should consider de-intensifying BP care in older Veterans has not been well-studied.
We aim to define Aggressive Hypertension Care (AHC) in Veterans age 65 and older using national VHA databases. First, we will validate data elements of AHC using medical record review. Next, we will test whether AHC (in comparison to adequate care) is associated with falls injury, and whether the risks outweigh reduction in strokes and cardiac events. Lastly, we will measure inter-facility variation in AHC. We will involve VHA providers throughout, to review results, guide analytic decisions and provide early identification of potential barriers to implementation. By the end of the award, we aim to develop a novel measure of appropriate hypertension care of relevance to older Veterans specific to age group, co-morbidity burden, and baseline risk for cardiovascular and fall events.
Data from the Corporate Data Warehouse (CDW) will be used to determine the presence of AHC (BP < 130/65 mmHg in combination with continuing 3+ or escalating 1+ BP medications) in primary care patients >65 with a diagnosis of hypertension between July 2011 and June 2013. A full review of the electronic health record will be conducted on a small subset of patients to substantiate the hypothesis that our data-based definition of AHC will achieve >85% specificity when compared to AHC determined by medical record review.
The CDW data will be merged with Medicare records to explore whether AHC is linked with increased risk of severe falls injury, and if so, whether the risks exceed the cardiovascular benefits (acute strokes and myocardial infarction). We will consider refinement of the AHC as guided by our provider panels and steering committee. We will consider subgroups of patients defined by baseline cardiovascular and falls risk.
Using the VA dataset, facility characteristics that predict AHC will be identified and the factors associated with the most variation between sites will be measured. We will consider results with provider panels and a steering committee to implement potential ways to reduce AHC.
Lastly, the panels will work with investigators to develop a performance measure to support appropriate hypertension care. Barriers to implementation will be identified, along with possible solutions.
Aim 1 - Validating data elements of AHC
We have completed chart review of 321 patients over 2 years of primary and cardiovascular care and have preliminary findings regarding the agreement between the chart review and the CDW. On preliminary analysis, in 3682 encounters and 8070 total medications in the CDW (the test) or chart review (the gold standard) there was a 2% false positive rate and a 14% false negative rate. With adjustment of the post-visit refill requirement (increasing from 124 days to 2 years) the false negative rate decreases to 10% and the false positive increases to 2.9%. We are unable to calculate sensitivity, specificity, or kappa, because there are no true positives (a medication cannot be neither CDW nor chart review) unless we add some determinant of what it means to not be on a medication. We are investigating possible strategies. There are plans for a paper on the agreement (false positives and negatives) in medication use between CDW and clinical notes, and further validation using medication count, SBP, AHC (dichotomous) as the measures. We will vary the frequency of encounters to count in the validation and consider inversely weighting the encounters within quarters. The chart abstraction data will also be used to identify the types of patients where what appears to be AHC is either 1) intentionally low BP to accommodate another condition (e.g. CHF); 2) the need to treat another condition with a drug that can cause low BP (furosemide for swelling); or 3) an acute condition that causes temporarily low BP (e.g. dehydration). This process will allow us to limit the CDW analysis to patients with actual AHC.
Analysis of interrater reliability continues.
Aim 2 - Harms and benefits of AHC
We will be linking VA data with Medicare data to measure the effect of AHC on net harms and benefits of AHC, and whether the net harms and benefits vary by patient complexity. The fall algorithm described above will be used as the fall injury outcome, along with standard ICD-9 algorithms for acute ischemic stroke, acute coronary syndrome, and acute heart failure, to measure net outcomes. We plan on testing both the dichotomous and continuous versions of AHC.
Work continues regarding how best to identify falls using medical record data. The manuscript "A Fall Injury Algorithm Developed in a National Administrative Healthcare Dataset: Balancing Accuracy with Inclusion" is nearing completion. It will present sequential algorithms for researchers and to use that identify fall injuries in administrative datasets across hospital, emergency room, nursing home, and outpatient data. Using a gold standard of patient interview and E-codes for fall injury in the Health and Retirement Study (HRS) merged with Medicare claims data, the manuscript describes how starting out with a core set of severe injuries in hospital and emergency room encounters results in algorithms that are more accurate (good PPV) however misses many of the reports of fall injury (sensitivity).
A decision has been made to identify stroke using the traditional primary diagnosis of acute ischemic and hemorrhagic strokes, and NOT the TIAs. MI is identified using primary and secondary diagnoses of MI, and not ACS. The work to date is based on VA hospitalizations using CDW data. The composite CVD outcome includes stroke (primary dx), MI and acute coronary syndrome (any dx), acute CHF (primary diagnosis only), and syncope. The 2-year incidence was 2.6% of the patient sample (1.3 million people). The next step is to merge with Medicare hospital data and capture the additional cases of exclusively non-VA hospitalizations.
Aim 3 - Facility-level variation in AHC
AHC is also being calculated at the facility level to determine the proportion of variation arising from the facility. Using a 2-level logistic regression (random effect of patient and facility), 68% of the overall variation in AHC is due to either patient or facility, and of this, only 1.7% is due to the facilities. However, the interfacility variation increases to almost 5% using continuous AHCc. Next, we will measure inter-facility variation as a continuous variable expressed as a function of the HDDs prescribed during each patient quarter and the patient's measured SBP. The next step will be to explore the impact of how many miles each patient lives from the nearest primary, secondary and tertiary sites.
Aim 4 - Assess barriers to implementation in the VA
The second meeting of the Leadership Panel was held in early December 2017. It appears that the physicians are more confident in the utility of using SPRINT finding to set goals for some patients. They find the new ACC/AHA Guidelines to be less helpful than they hoped because of too many decision points and how difficult it is to get an accurate BP reading. All endorse changes in infrastructure and process that would better mirror the measurement method used by SPRINT.
The panel members had several ideas for facility level variables that might be useful. These include the mean age of PC providers, the proportion of HTN care under the jurisdiction of PC, proportion of patients getting care outside the VA, what a facility's performance is on non-HTN measures and how rural the facility is.
Our next panel discussion is planned for January 2019.
The chart abstraction will add to the existing work on the accuracy of VA administrative data.
This project attempts to balance geriatric outcomes, such as falls and fall injury, with traditional cardiovascular outcomes. It will support the use of individual patient data to inform decisions about net benefit (or harm) of various treatment strategies. This research fits clearly with VHA priorities to reduce use of inappropriate and harmful treatments and to enhance patient-centered care. The research has the possibility of informing future performance measures and decision support tools to help minimize both cardiovascular and fall injury, and appropriately individualize care for older adults with hypertension.
External Links for this Project
Grant Number: I01HX001611-01A2
- Min L, Ha JK, Aubert CE, Hofer TP, Sussman JB, Langa KM, Tinetti M, Kim HM, Maciejewski ML, Gillon L, Larkin A, Chan CL, Kerr EA, Bravata D, Cushman WC. A Method to Quantify Mean Hypertension Treatment Daily Dose Intensity Using Health Care System Data. JAMA Network Open. 2021 Jan 4; 4(1):e2034059. [view]
- Aubert CE, Ha JK, Kim HM, Rodondi N, Kerr EA, Hofer TP, Min L. Clinical outcomes of modifying hypertension treatment intensity in older adults treated to low blood pressure. Journal of the American Geriatrics Society. 2021 Oct 1; 69(10):2831-2841. [view]
- Aubert CE, Chan CL, Terman SW, Hofer TP, Ha JK, Cushman WC, Sussman J, Min L. Evaluating alternative methods of comparing antihypertensive treatment intensity. The American journal of managed care. 2022 May 1; 28(5):e157-e162. [view]
- Aubert CE, Ha JK, Kerr EA, Hofer TP, Min L. Factors associated with antihypertensive treatment intensification and deintensification in older outpatients. International Journal of Cardiology. Hypertension. 2021 Jun 1; 9:100098. [view]
- Min L, Tinetti M, Langa KM, Ha J, Alexander N, Hoffman GJ. Measurement of Fall Injury With Health Care System Data and Assessment of Inclusiveness and Validity of Measurement Models. JAMA Network Open. 2019 Aug 2; 2(8):e199679. [view]
- Min L, Ha JK, Hofer TP, Sussman J, Langa K, Cushman WC, Tinetti M, Kim HM, Maciejewski ML, Gillon L, Larkin A, Chan CL, Kerr E. Validation of a Health System Measure to Capture Intensive Medication Treatment of Hypertension in the Veterans Health Administration. JAMA Network Open. 2020 Jul 1; 3(7):e205417. [view]